Notes and sources
All 138 notes from the book, by chapter, keyed to the phrase they support, with a live link to each source. Where a source is vendor research, the note says what the vendor sells.
Chapter 1The Unprecedented Moment23 notes
- 1
Barry Staw called it “escalation of commitment”
Barry M. Staw, “Knee-Deep in the Big Muddy: A Study of Escalating Commitment to a Chosen Course of Action,” Organizational Behavior and Human Performance 16, no. 1 (1976): 27–44; Barry M. Staw and Jerry Ross, “Understanding Behavior in Escalation Situations,” Science 246, no. 4927 (1989): 216–20. Staw’s 1976 experiment located the effect in personal responsibility and self-justification; Staw and Ross (1989) group its causes as project, psychological, social and organisational.
doi.org/10.1016/0030-5073(76)9000… ↗doi.org/10.1126/science.246.4927.… ↗
- 2
Somewhere between 30 and 40 per cent of software projects
Mark Keil, Joan Mann, and Arun Rai, “Why Software Projects Escalate: An Empirical Analysis and Test of Four Theoretical Models,” MIS Quarterly 24, no. 4 (2000): 631–47. Keil, Mann and Rai’s estimate, for information systems projects.
- 3
Microsoft’s own telemetry
Vasu Jakkal, “80% of Fortune 500 Use Active AI Agents: Observability, Governance, and Security Shape the New Frontier,” Microsoft Security Blog, 10 February 2026. Microsoft’s own telemetry: agents built with Copilot Studio or Agent Builder that were deployed and active in the last 28 days of November 2025.
- 4
Stanford’s 2026 AI Index
Stanford Institute for Human-Centered Artificial Intelligence, “Economy,” in The 2026 AI Index Report (Stanford, CA: Stanford University, 2026); Stanford Institute for Human-Centered Artificial Intelligence, “Economy,” in The 2025 AI Index Report (Stanford, CA: Stanford University, 2025). The 2025 Index gives the 78 per cent for the year before.
hai.stanford.edu/ai-index/2026-ai-index-re… ↗hai.stanford.edu/ai-index/2025-ai-index-re… ↗
- 5
GTG-1002
Anthropic, “Disrupting the First Reported AI-Orchestrated Cyber Espionage Campaign,” November 2025; summary post “Disrupting an AI-Orchestrated Cyber Espionage Campaign,” Anthropic, 13 November 2025. The account is Anthropic’s own investigation of misuse of its own product.
- 6
several respected researchers argued
Toby Murray, “An AI Lab Says Chinese-Backed Bots Are Running Cyber Espionage Attacks. Experts Have Questions,” The Conversation, 17 November 2025.
- 7
McKinsey’s 2026 AI Trust Maturity Survey
McKinsey & Company, “State of AI Trust in 2026: Shifting to the Agentic Era,” Tech Forward (blog), 25 March 2026. About 500 organisations, surveyed December 2025 to January 2026.
- 8
State of AI in Business
Aditya Challapally, Chris Pease, Ramesh Raskar, and Pradyumna Chari, “The GenAI Divide: State of AI in Business 2025” (MIT NANDA, July 2025). The report describes its method as a review of more than 300 publicly disclosed AI initiatives, structured interviews with representatives of 52 organisations and survey responses from 153 senior leaders, and labels itself preliminary findings.
- 9
Global Cybersecurity Outlook 2026
World Economic Forum, Global Cybersecurity Outlook 2026, Insight Report, in collaboration with Accenture (Geneva: World Economic Forum, January 2026). The 804 respondents include 316 CISOs, 105 chief executives and 123 other C-suite executives.
- 10
Proofpoint’s 2026 research
Proofpoint, “Proofpoint Research Reveals Half of Global Organizations Experienced AI Incidents Despite Having AI Security Controls in Place,” press release, 28 April 2026 (on the 2026 AI and Human Risk Landscape report). Survey of more than 1,400 security professionals in 12 countries. The figure is for a confirmed or suspected AI-related incident. Proofpoint sells email, collaboration and AI security products.
- 11
Nutanix’s 2026 Enterprise Cloud Index
Nutanix, “Nutanix Enterprise Cloud Index Finds AI Is Driving Rapid Container Adoption while Shadow IT and Organizational Silos Create AI Risks,” press release, 3 March 2026. Wakefield Research survey of 1,600 IT and engineering decision-makers, November 2025. Nutanix sells hybrid-cloud infrastructure.
- 12
for high-risk systems, the deployer
Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act), Official Journal of the European Union L, 12 July 2024, arts. 2(1)(c), 26, 99. Article 26 sets these duties for deployers of high-risk AI systems; Article 2(1)(c) extends the Act to providers and deployers outside the EU where the system’s output is used in the EU; Article 99 sets the fines.
- 13
has already moved once
Regulation (EU) 2026/1744 of the European Parliament and of the Council of 8 July 2026 amending Regulations (EU) 2024/1689, (EU) 2018/1139 and (EU) 2023/1230 as regards the simplification of the implementation of harmonised rules on artificial intelligence (Digital Omnibus on AI), Official Journal of the European Union L, 24 July 2026. Annex III high-risk obligations now apply from 2 December 2027, and Annex I obligations from 2 August 2028.
- 14
most states have passed AI laws
Scott Babwah Brennen, “Where State AI Legislation Stands Half Way Into 2026,” Tech Policy Press, 6 July 2026. Tech Policy Press counts 29 states enacting AI legislation in the first half of 2026, and 46 over 2025.
- 15
The UK went the other way
Department for Science, Innovation and Technology, A Pro-Innovation Approach to AI Regulation: Government Response (London: Department for Science, Innovation and Technology, 6 February 2024).
- 16
AI Governance Guidelines
Press Information Bureau, Government of India, “India AI Governance Guidelines: Enabling Safe and Trusted AI Innovation,” backgrounder, 15 February 2026.
- 17
Seventeen African countries
OECD, “AI Governance in Africa: Insights from a Policy Dialogue with 12 Countries,” OECD Artificial Intelligence Case Studies (Paris: OECD Publishing, 30 April 2026); African Union, Continental Artificial Intelligence Strategy: Harnessing AI for Africa’s Development and Prosperity (Addis Ababa: African Union, July 2024). The OECD’s table lists 17 African countries adopting national AI strategies from 2018 to April 2026, 13 of them since 2023.
oecd.org/content/dam/oecd/en/publi… ↗au.int/en/documents/20240809/con… ↗
- 18
On 15 July 2026, the government
Anthony Albanese, Tim Ayres, and Andrew Charlton, “AI in Australia’s Interests,” media release, Prime Minister of Australia, 15 July 2026. The standards announced are rules for large data centres and AI training, building on the Data Centre Expectations, and are expected to be legislated in 2027.
- 19
its National AI Plan
Department of Industry, Science and Resources, National AI Plan (Canberra: Australian Government, 2 December 2025); Department of Industry, Science and Resources, “Keep Australians Safe,” in National AI Plan (Canberra: Australian Government, 2 December 2025).
industry.gov.au/publications/national-ai-… ↗industry.gov.au/publications/national-ai-… ↗
- 20
REP 798
Australian Securities and Investments Commission, Beware the Gap: Governance Arrangements in the Face of AI Innovation, Report 798 (Sydney: ASIC, October 2024); Australian Securities and Investments Commission, “24-238MR ASIC Warns Governance Gap Could Emerge in First Report on AI Adoption by Licensees,” media release, October 2024. ASIC reviewed 624 AI use cases at 23 licensees. Its words: “there is the potential for a governance gap – one that risks widening if AI adoption outpaces governance in response to competitive pressures.” The National AI Plan followed on 2 December 2025.
download.asic.gov.au/media/mtllqjo0/rep-798-pu… ↗asic.gov.au/about-asic/news-centre/fi… ↗
- 21
signed by Therese McCarthy Hockey
Therese McCarthy Hockey, “APRA Letter to Industry on Artificial Intelligence (AI),” letter to all APRA-regulated entities, Australian Prudential Regulation Authority, 30 April 2026. The letter reports AI tools used outside approved control frameworks and weak controls over post-deployment monitoring and decommissioning; an inventory of AI use and clear lifecycle ownership appear among the minimum arrangements APRA expects.
- 22
coordinated call to action on frontier AI
Australian Securities and Investments Commission, “26-201MR ASIC and APRA Warn Frontier AI Awareness Must Turn to Action,” media release, 27 August 2026; Australian Prudential Regulation Authority and Australian Securities and Investments Commission, Resilience at Frontier AI Speed: Insights from the APRA-ASIC Industry Roundtables, joint publication, August 2026. The nine roundtables ran in June and July 2026; the joint release and paper followed on 27 August.
asic.gov.au/about-asic/news-centre/fi… ↗download.asic.gov.au/media/imunrchk/asic-and-a… ↗
- 23
AI skills sit at the top of the global shortage list
ManpowerGroup, “Global Talent Shortage Reaches Turning Point as AI Skills Claim Top Spot,” press release, 26 February 2026; Gartner, “Gartner Survey Finds AI Automation Is Reducing Some Entry Level Hiring at Nearly One-Quarter of Organizations,” press release, Stamford, CT, 27 July 2026. ManpowerGroup: 72 per cent of 39,000 employers in 41 countries report difficulty filling roles, and AI model and application development (20 per cent) and AI literacy (19 per cent) lead the hardest-to-find skills. Gartner: 22 per cent of 110 heads of HR report that at least one business leader has stopped hiring for entry-level roles because of AI automation.
manpowergroup.com/en/news-releases/news/glo… ↗gartner.com/en/newsroom/press-release… ↗
Chapter 2Your Expertise Is Your Superpower17 notes
- 1
The World Economic Forum’s projection
World Economic Forum, Future of Jobs Report 2025, Insight Report (Geneva: World Economic Forum, January 2025). The WEF projects 92 million jobs displaced and 170 million created by 2030, a net gain of 78 million, across all the macrotrends it studies, not automation alone.
- 2
GDPR fines of up to 4 per cent
Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation), art. 83(5), OJ L 119, 4.5.2016. The 4 per cent is of total worldwide annual turnover for the preceding financial year.
- 3
Leaked AWS keys
William Gamazo and Nathaniel Quist, “CloudKeys in the Air: Tracking Malicious Operations of Exposed IAM Keys,” Unit 42 (Palo Alto Networks), 30 October 2023; Paul Kunert, “DXC Spills AWS Private Keys on Public GitHub,” The Register, 14 November 2017. Unit 42 saw exposed keys used within five minutes, for cryptomining; in the DXC case the bill reached US$64,000 over four days.
unit42.paloaltonetworks.com/malicious-operations-of-e… ↗theregister.com/2017/11/14/dxc_github_aws… ↗
- 4
Gary Klein
Gary A. Klein, Roberta Calderwood, and Anne Clinton-Cirocco, “Rapid Decision Making on the Fire Ground,” Proceedings of the Human Factors Society Annual Meeting 30, no. 6 (1986): 576–580; Gary A. Klein, Roberta Calderwood, and Anne Clinton-Cirocco, Rapid Decision Making on the Fire Ground, ARI Technical Report 796 (Alexandria, VA: U.S. Army Research Institute for the Behavioral and Social Sciences, June 1988); Gary Klein, Sources of Power: How People Make Decisions (Cambridge, MA: MIT Press, 1998). In the study’s example the commander sensed that a roof was about to collapse.
doi.org/10.1177/154193128603000616 ↗apps.dtic.mil/sti/citations/ADA199492 ↗
- 5
we can know more than we can tell
Michael Polanyi, The Tacit Dimension (Garden City, NY: Doubleday, 1966), 4.
- 6
Michelle Vaccaro, Abdullah Almaatouq and Thomas Malone
Michelle Vaccaro, Abdullah Almaatouq, and Thomas Malone, “When Combinations of Humans and AI Are Useful: A Systematic Review and Meta-analysis,” Nature Human Behaviour 8 (2024): 2293–2303.
- 7
Dell’Acqua
Fabrizio Dell’Acqua, Edward McFowland III, Ethan Mollick, Hila Lifshitz-Assaf, Katherine C. Kellogg, Saran Rajendran, Lisa Krayer, François Candelon, and Karim R. Lakhani, “Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality,” Harvard Business School Working Paper 24-013 (Harvard Business School, September 2023). The 758 consultants were each assigned to one of two experiments, one inside the frontier and one outside it.
- 8
APT36
Radu Tudorica, “APT36: A Nightmare of Vibeware,” Bitdefender Business Insights (blog), 5 March 2026. Bitdefender describes a Pakistan-based actor rewriting payloads in languages including Nim, Zig, Crystal, Rust and Go; it gives no count of languages. Bitdefender sells endpoint detection and response products whose behavioural detection the report recommends.
- 9
ENISA’s 2025 threat report
European Union Agency for Cybersecurity (ENISA), ENISA Threat Landscape 2025 (Athens: ENISA, October 2025; version 1.3, 22 September 2026); KnowBe4, Phishing Threat Trends Report 2025, KnowBe4, 2025. ENISA relays the figure from KnowBe4; its own body text says over 80 per cent of phishing emails identified between September 2024 and February 2025 used AI to some extent. KnowBe4 sells security-awareness training and AI-driven email security products.
doi.org/10.2824/1946374 ↗knowbe4.com/hubfs/Phishing-Threat-Tre… ↗
- 10
State of Exploitation 2026
Patrick Garrity, “VulnCheck State of Exploitation 2026,” VulnCheck (blog), 21 January 2026. VulnCheck sells exploit and vulnerability intelligence, including the VulnCheck KEV catalogue this analysis is drawn from.
- 11
minus seven days
Jurgen Kutscher, “M-Trends 2026: Data, Insights, and Strategies From the Frontlines,” Google Cloud Blog, 24 March 2026; Casey Charrier and Robert Weiner, “How Low Can You Go? An Analysis of 2023 Time-to-Exploit Trends,” Google Cloud Blog (Mandiant), 16 October 2024. Mandiant measures time to exploit against the release of a patch, and describes the −7 days as an estimated mean.
cloud.google.com/blog/topics/threat-intell… ↗cloud.google.com/blog/topics/threat-intell… ↗
- 12
put the global workforce gap at 4.8 million
ISC2, “Growth of Cybersecurity Workforce Slows in 2024 as Economic Uncertainty Persists,” press release, Alexandria, VA, 11 September 2024; ISC2, “2025 ISC2 Cybersecurity Workforce Study,” ISC2 Insights, 4 December 2025. ISC2 says it dropped the estimate because respondents now rank missing skills above missing headcount.
isc2.org/Insights/2024/09/ISC2-Pub… ↗isc2.org/Insights/2025/12/2025-ISC… ↗
- 13
Klarna announced
Klarna, “Klarna AI Assistant Handles Two-Thirds of Customer Service Chats in Its First Month,” press release, New York, 27 February 2024; Charles Daly, “Klarna Turns From AI to Real Person Customer Service,” Bloomberg, 8 May 2025; Rhys Fisher, “Klarna Redeploys Staff to Customer Service, Pulls From Marketing, Engineering, and Legal Teams,” CX Today, 3 September 2025. Klarna’s 2024 release put the assistant’s customer satisfaction on par with human agents. The staff redeployment was reported in September 2025 and named marketing, engineering and legal staff.
klarna.com/international/press/klarn… ↗bloomberg.com/news/articles/2025-05-08/… ↗cxtoday.com/contact-center/klarna-red… ↗
- 14
delivered a report to an Australian government department
Deloitte, Targeted Compliance Framework: Final Report (Canberra: Department of Employment and Workplace Relations, 26 September 2025; replaces report dated 4 July 2025); Krishani Dhanji, “Deloitte to Pay Money Back to Albanese Government after Using AI in $440,000 Report,” Guardian (Australia), 6 October 2025; Carly Page, “Deloitte Refunds Aussie Gov after AI Fabrications Slip into $440K Welfare Report,” The Register, 6 October 2025. The report is dated 4 July 2025; the errors were reported in August, and the corrected version of 26 September 2025 discloses use of a generative AI tool chain in part of the analysis.
dewr.gov.au/assuring-integrity-target… ↗theguardian.com/australia-news/2025/oct/0… ↗theregister.com/2025/10/06/deloitte_ai_re… ↗
- 15
DryRun Security
DryRun Security, The Agentic Coding Security Report (Austin, TX: DryRun Security, March 2026); announced in DryRun Security, “New DryRun Security Research: Anthropic’s Claude Generates the Most Unresolved Security Flaws in AI-Built Applications,” GlobeNewswire, 11 March 2026. DryRun Security sells AI code-security review, the kind of tool the report recommends and used to score the results.
- 16
Hosseini Maasoum and Lichtinger
Seyed Mahdi Hosseini Maasoum and Guy Lichtinger, “Generative AI as Seniority-Biased Technological Change: Evidence from U.S. Résumé and Job Posting Data” (working paper, Harvard University, first version 31 August 2025; version of 27 October 2025), SSRN. The paper has since been revised (June 2026) to cover 65 million workers; the 9 per cent is the fall in junior employment across adopting firms, concentrated in high-exposure roles, and it measures senior employment rather than hiring.
- 17
Other researchers reading the same period disagree
Richard Audoly, Miles Guerin, and Giorgio Topa, “Do Job Postings Show Early Labor-Market Effects of AI?,” Liberty Street Economics (Federal Reserve Bank of New York), 14 May 2026.
Chapter 3The AI Layer23 notes
- 1
developers using GitHub Copilot completed
Sida Peng, Eirini Kalliamvakou, Peter Cihon, and Mert Demirer, “The Impact of AI on Developer Productivity: Evidence from GitHub Copilot,” preprint, arXiv, 13 February 2023. GitHub sells Copilot. The measured figure is 55.8 per cent.
- 2
METR
Joel Becker, Nate Rush, Beth Barnes, and David Rein, “Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity,” METR, 10 July 2025; Joel Becker, Nate Rush, Tom Cunningham, David Rein, and Khalid Mahamud, “We are Changing our Developer Productivity Experiment Design,” METR, 24 February 2026. METR says it is changing its study design, citing developers declining to work without AI, selection effects and a lower pay rate, and that these affect a minority of developers and tasks.
metr.org/blog/2025-07-10-early-202… ↗metr.org/blog/2026-02-24-uplift-up… ↗
- 3
close to every job posting in the United States
Ana Elena Azpúrua, “Enhance or Eliminate? How AI Will Likely Change These Jobs,” HBS Working Knowledge, 20 February 2026; Wilbur Xinyuan Chen, Suraj Srinivasan, and Saleh Zakerinia, “Displacement or Complementarity? The Labor Market Impact of Generative AI,” Harvard Business School Working Paper 25-039, December 2024. The 13 and 20 per cent figures are from the August 2025 update, as reported by HBS Working Knowledge; the December 2024 version gives 17 and 22 per cent. Srinivasan is at HBS, his co-authors at HKUST and Ohio State.
library.hbs.edu/working-knowledge/enhance… ↗hbs.edu/ris/download.aspx?name=25… ↗
- 4
Friedrich Hayek
F. A. Hayek, “The Use of Knowledge in Society,” American Economic Review 35, no. 4 (September 1945): 519–30.
- 5
a system called IRIS
Ziyang Li, Saikat Dutta, and Mayur Naik, “IRIS: LLM-Assisted Static Analysis for Detecting Security Vulnerabilities,” conference paper, International Conference on Learning Representations (ICLR 2025); preprint, arXiv:2405.17238, version 3, 6 April 2025. Academic research from the University of Pennsylvania and Cornell University, supported by US National Science Foundation award CCF 2313010; no product is sold. The model was GPT-4, used to infer which library calls bring untrusted data in and which are dangerous to pass it to; CodeQL ran the analysis. The 27 and 55 are out of the 120 vulnerabilities in the authors’ CWE-Bench-Java dataset; the four previously unknown vulnerabilities came from a separate run on the latest versions of 30 Java projects. The false discovery rate fell by about five percentage points against CodeQL alone. The authors treat that rate as an upper bound; their manual review of 50 randomly sampled alarms found 27 that showed a potential attack surface, a false discovery rate of about 46 per cent. An earlier version of the paper reported six.
- 6
only 56 per cent of generations were secure
Veracode, 2026 GenAI Code Security Report (Burlington, MA: Veracode, 2026); Veracode, “2026 GenAI Code Security: Syntax is Solved, Security is Not,” Veracode Blog, 28 July 2026; Felix Brombacher, “Spring 2026 GenAI Code Security Update: Despite Claims, AI Models Are Still Failing Security,” Veracode Blog, 24 March 2026; Jens Wessling, “We Asked 100+ AI Models to Write Code. Here’s How Many Failed Security Tests,” Veracode Blog, 30 July 2025; Veracode, “October 2025 Update: GenAI Code Security Report,” Veracode, October 2025. The four runs are July 2025, October 2025, March 2026 and July 2026. Veracode sells application-security testing.
veracode.com/resources/analyst-reports… ↗veracode.com/blog/2026-genai-code-secu… ↗veracode.com/blog/spring-2026-genai-co… ↗veracode.com/blog/genai-code-security-… ↗veracode.com/resources/analyst-reports… ↗
- 7
PCI DSS
PCI Security Standards Council, “Securing the Future of Payments: PCI SSC Publishes PCI Data Security Standard v4.0,” press release, 31 March 2022.
- 8
114 controls
International Organization for Standardization and International Electrotechnical Commission, ISO/IEC 27001:2022, Information Security, Cybersecurity and Privacy Protection – Information Security Management Systems – Requirements, 3rd ed. (Geneva: ISO, 2022); PECB, “ISO/IEC 27002:2022 – Information Security, Cybersecurity, and Privacy Protection,” PECB whitepaper; Danny Manimbo, “The New ISO 27002:2022 - What You Need to Know,” Schellman, 4 August 2023. Of the 93 controls, 11 are new and 57 earlier controls were merged into 24.
iso.org/standard/27001 ↗pecb.com/en/whitepaper/isoiec-2700… ↗schellman.com/blog/iso-certifications/i… ↗
- 9
internal audit and a documented management review
ISO/IEC 27001:2022, clause 9.2.1, as quoted in High Table, “ISO 27001 Clause 9.2 Internal Audit”; ISO/IEC 27001:2022, clause 9.3, as quoted in High Table, “ISO 27001 Clause 9.3 Management Review”; National Institute of Standards and Technology, The NIST Cybersecurity Framework (CSF) 2.0, NIST CSWP 29 (Gaithersburg, MD: NIST, 26 February 2024). CSF 2.0 does include oversight outcomes (GV.OV) under which strategy and performance are reviewed, without prescribing an audit program or scheduled management review.
hightable.io/iso-27001-clause-9-2-inte… ↗hightable.io/iso-27001-clause-9-3-mana… ↗doi.org/10.6028/NIST.CSWP.29 ↗
- 10
A sixth function, Govern
National Institute of Standards and Technology, “NIST Releases Version 2.0 of Landmark Cybersecurity Framework,” news release, 26 February 2024; National Institute of Standards and Technology, Framework for Improving Critical Infrastructure Cybersecurity, Version 1.1 (Gaithersburg, MD: NIST, 16 April 2018). CSF 2.0 succeeded version 1.1 (16 April 2018), which carried a Governance category (ID.GV) under Identify.
nist.gov/news-events/news/2024/02/… ↗doi.org/10.6028/NIST.CSWP.04162018 ↗
- 11
Cyentia Institute
Wade Baker, “Prioritization to Prediction, Vol 3: Wade’s Take,” Cyentia Institute, 12 March 2019; Kenna Security, “More Resources Don’t Lead to Better or Faster Vulnerability Management, Kenna Security Report Finds,” press release, GlobeNewswire, 12 March 2019; Kenna Security and Cyentia Institute, Prioritization to Prediction, Volume 4: Measuring What Matters in Remediation (Kenna Security, 2019). The series was commissioned by Kenna Security, which sells vulnerability prioritisation. Volume 4 found that patch tooling and budget do raise remediation capacity.
cyentia.com/p2p-vol3-wade ↗globenewswire.com/news-release/2019/03/12/1… ↗dokumen.pub/prioritization-to-predict… ↗
- 12
Later work in the same series
Kenna Security and Cyentia Institute, Prioritization to Prediction, Volume 7: Establishing Defender Advantage (Kenna Security, 2021); Jay Jacobs, Sasha Romanosky, Octavian Suciu, Benjamin Edwards, and Armin Sarabi, “Enhancing Vulnerability Prioritization: Data-Driven Exploit Predictions with Community-Driven Insights,” preprint, arXiv, 27 February 2023 (rev. 15 June 2023). Volume 7 draws on scan and remediation data from nearly 500 organisations and over 6 billion vulnerabilities, but its analysis is of exploit timing; the later remediation-capacity figure, reported by Jacobs and colleagues, is a monthly median of 15.5 per cent, varying widely between organisations. Kenna Security sells risk-based vulnerability management software.
images.g2crowd.com/uploads/attachment/file/1… ↗arxiv.org/abs/2302.14172 ↗
- 13
11.3 per cent
Natalie Tischler, “The 36% Surge in High-Risk Vulnerabilities: What It Means for Your Business,” Veracode Blog, 10 March 2026 (reporting Veracode, 2026 State of Software Security); Veracode, “Veracode 2026 State of Software Security Report Reveals Four Out of Five Organizations Are Drowning in Security Debt,” press release, Business Wire, 24 February 2026, reprinted in The AI Journal. Veracode sells application-security testing.
veracode.com/blog/high-risk-vulnerabil… ↗aijourn.com/veracode-2026-state-of-so… ↗
- 14
Gartner sells advisory subscriptions
Rajesh Kandaswamy, “Be AI-First to Outperform Your Competition by 2028,” Gartner, 13 November 2025. Gartner sells research and advisory subscriptions on AI strategy.
- 15
Ponemon
Ponemon Institute, “Why We Are Unique,” Ponemon Institute; Ponemon Institute, “Ponemon Library”.
ponemon.org/about/why-we-are-unique.h… ↗ponemon.org/research/ponemon-library ↗
- 16
CVE-2025-48757
MITRE, “CVE-2025-48757,” CVE Record, published 30 May 2025; Matt Palmer, “Statement on CVE-2025-48757,” 29 May 2025. The CVE record and the researcher place the flaw in Lovable; Lovable disputes it. The researcher counts 170 affected apps with row-level security missing or inadequate.
cve.org/CVERecord?id=CVE-2025-48757 ↗mattpalmer.io/posts/statement-on-CVE-20… ↗
- 17
Microsoft Power Apps
UpGuard Team, “By Design: How Default Permissions on Microsoft Power Apps Exposed Millions,” UpGuard, 23 August 2021.
- 18
IBM’s Cost of a Data Breach research put
IBM, Cost of a Data Breach Report 2025: The AI Oversight Gap (Armonk, NY: IBM, 2025). The 2025 global average fell 9 per cent to US$4.44 million, the first decline in five years.
- 19
OX Security
Moshe Siman Tov Bustan, Mustafa Naamnih, Nir Zadok, and Roni Bar, “The Mother of All AI Supply Chains: Critical, Systemic Vulnerability at the Core of Anthropic’s MCP,” OX Security, 15 April 2026; Cloud Security Alliance AI Safety Initiative, “MCP by Design: RCE Across the AI Agent Ecosystem,” CSA research note, 20 April 2026. OX reports the flaw across the official MCP SDKs and up to 200,000 exposed instances; Anthropic, the protocol’s maintainer, called the behaviour expected. OX Security sells application and AI supply-chain security.
ox.security/blog/the-mother-of-all-ai… ↗labs.cloudsecurityalliance.org/research/csa-research-not… ↗
- 20
OpenAI models escaped a sandboxed
OpenAI, “OpenAI and Hugging Face Partner to Address Security Incident during Model Evaluation,” OpenAI, 21 July 2026; OpenAI, “The Hugging Face Incident and the Road Ahead,” OpenAI, 26 August 2026; Hugging Face, “Security Incident Disclosure – July 2026,” Hugging Face Blog, 16 July 2026. OpenAI attributes the incident to OpenAI models “including GPT‑5.6 Sol and an even more capable pre-release model”; the breach reached part of Hugging Face’s production infrastructure.
openai.com/index/hugging-face-model-… ↗openai.com/index/hugging-face-incide… ↗huggingface.co/blog/security-incident-ju… ↗
- 21
ISO/IEC 42001
International Organization for Standardization and International Electrotechnical Commission, ISO/IEC 42001:2023, Information Technology – Artificial Intelligence – Management System, 1st ed. (Geneva: ISO, 2023); Advisera, “ISO 42001 Requirements: Clauses and Structure”; Cyberzoni, “ISO/IEC 42001 Clause 8 Operation – Guidance & Best Practices”. Clause and Annex A details are from practitioner summaries; the standard’s text is licensed.
iso.org/standard/42001 ↗advisera.com/articles/iso-42001-clause… ↗cyberzoni.com/standards/iso-42001/claus… ↗
- 22
Julian Rotter
Julian B. Rotter, “Generalized Expectancies for Internal versus External Control of Reinforcement,” Psychological Monographs: General and Applied 80, no. 1 (1966): 1–28.
- 23
Csikszentmihalyi
Fausto Massimini, Mihaly Csikszentmihalyi, and Massimo Carli, “The Monitoring of Optimal Experience: A Tool for Psychiatric Rehabilitation,” Journal of Nervous and Mental Disease 175, no. 9 (1987): 545–49.
Chapter 4Thinking Clearly Under Threat11 notes
- 1
LSA Global, a consultancy
LSA Global, “LSA 3x Organizational Alignment Model”; LSA Global, “The Power of Strategic Alignment,” LSA Global blog. LSA gives no survey date or sample design. LSA Global sells organisational alignment consulting built on this model.
- 2
settled charges in two cases
US Securities and Exchange Commission, “SEC Charges Two Investment Advisers with Making False and Misleading Statements About Their Use of Artificial Intelligence,” press release 2024-36, 18 March 2024. The two firms settled without admitting or denying the findings.
- 3
Herbert Clark
Herbert H. Clark and Susan E. Brennan, “Grounding in Communication,” in Perspectives on Socially Shared Cognition, ed. Lauren B. Resnick, John M. Levine and Stephanie D. Teasley (Washington, DC: American Psychological Association, 1991), 127–49.
- 4
Nine days earlier, OpenAI had disclosed
OpenAI, “OpenAI and Hugging Face Partner to Address Security Incident during Model Evaluation,” OpenAI, 21 July 2026; Hugging Face, “Security Incident Disclosure – July 2026,” Hugging Face blog, 16 July 2026.
openai.com/index/hugging-face-model-… ↗huggingface.co/blog/security-incident-ju… ↗
- 5
On 30 July 2026, Anthropic published
Anthropic, “Investigating Three Real-World Incidents in Our Cybersecurity Evaluations,” Anthropic Research, 30 July 2026; Anthropic, “An Alignment Assessment of Recent Cybersecurity Incidents,” Anthropic Research, 9 September 2026. In its 9 September follow-up, Anthropic revised its July view that the incidents were closer to operational failures, concluding that Claude’s behaviour reflected biased reasoning and recklessness.
anthropic.com/research/investigating-in… ↗anthropic.com/research/alignment-assess… ↗
- 6
calls it “excessive agency”
OWASP GenAI Security Project, “Preface,” OWASP Top 10 for LLM Applications 2026, GitHub; OWASP GenAI Security Project, “OWASP Top 10 for LLM Applications 2025”. Excessive agency was LLM06 in the 2025 edition.
raw.githubusercontent.com/GenAI-Security-Project/Ge… ↗genai.owasp.org/llm-top-10 ↗
- 7
Kiro
Robert Hart, “Amazon Blames Human Employees for an AI Coding Agent’s Mistake,” The Verge, 20 February 2026, reporting the Financial Times; Amazon Staff, “Correcting the Financial Times Report about AWS, Kiro, and AI,” About Amazon, 20 February 2026; AI Incident Database, “Incident 1442: Kiro AI Coding Tool Was Reportedly Implicated in 13-Hour AWS Cost Explorer Outage in Mainland China.” The account comes from anonymous sources speaking to the Financial Times; Amazon attributes the interruption to user error, specifically misconfigured access controls.
- 8
Top 10 for Agentic Applications
OWASP GenAI Security Project, OWASP Top 10 for Agentic Applications for 2026 (OWASP Foundation, 9 December 2025).
- 9
Lisanne Bainbridge
Lisanne Bainbridge, “Ironies of Automation,” Automatica 19, no. 6 (1983): 775–79.
- 10
cognitive sovereignty
See, for example, Robert Atkinson, “Cognitive Sovereignty and Neurocomputational Harm in Predictive Digital Platforms,” Ethics and Information Technology 27, no. 4 (2025): 66; Chia-Wei Sun, “Residual Governance for Responsible Embodied AI: Authority, Contestability, and Cognitive Sovereignty,” AI and Ethics 6, no. 5 (2026): 555. The term is used in several senses, including in work on neurotechnology.
doi.org/10.1007/s10676-025-09873-y ↗doi.org/10.1007/s43681-026-01400-z ↗
- 11
John Keats
John Keats to George and Thomas Keats, 22 December 1817, in Letters of John Keats to His Family and Friends, ed. Sidney Colvin (Project Gutenberg, 2011). Some editions date the letter 21 December.
Chapter 5The Market You Cannot See25 notes
- 1
nearly triple by 2034
Fortune Business Insights, “Cybersecurity Market Size, Share & Industry Analysis, By Component (Solutions and Services), By Deployment (On-premises and Cloud), By Security Type ..., By Enterprise Size ..., By Industry ..., and Regional Forecast, 2026–2034,” report FBI101165, last updated 29 June 2026. Fortune Business Insights sells market-research reports, including this one.
- 2
According to IBM’s 2025 Cost of a Data Breach Report
IBM Security, Cost of a Data Breach Report 2025 (Armonk, NY: IBM, 2025). Ponemon Institute research for IBM on organisations breached between March 2024 and February 2025. The saving is for security teams using AI and automation extensively, against those that did not.
- 3
Healthcare has sat at the top of IBM’s table
IBM Security, Cost of a Data Breach Report 2025 (Armonk, NY: IBM, 2025). IBM reports healthcare as the highest-cost industry for the 12th consecutive year.
- 4
Mandiant’s median dwell time
Jurgen Kutscher, “M-Trends 2026: Data, Insights, and Strategies From the Frontlines,” Google Cloud Blog, 23 March 2026. Global median dwell time rose to 14 days from 11.
- 5
In May 2024, Ticketmaster discovered
Ticketmaster LLC, “Notice of Data Breach” (sample consumer notice [T01] US-General), submitted to the California Office of the Attorney General, 2024; State of California Department of Justice, Office of the Attorney General, “Submitted Breach Notification Sample: Ticketmaster LLC”; Office of the Maine Attorney General, “Data Breach Notifications: Ticketmaster LLC,” filed July 2024; Live Nation Entertainment, Inc., Form 8-K, Current Report (Date of Earliest Event Reported: 20 May 2024), filed 31 May 2024, US Securities and Exchange Commission; Jonathan Greig, “Live Nation Confirms Ticketmaster Breach after Hackers Hawk Stolen Info of 560 Million,” The Record from Recorded Future News, 3 June 2024. Live Nation’s filing gives 20 May as the date it identified the activity, in a cloud database hosted by a third-party data services provider, and 23 May as the date it found personal data may have been affected.
oag.ca.gov/system/files/[T01] US-Gen… ↗oag.ca.gov/ecrime/databreach/reports… ↗maine.gov/agviewer/content/ag/98523… ↗sec.gov/Archives/edgar/data/13352… ↗therecord.media/live-nation-confirms-tick… ↗
- 6
Mandiant found credentials were stolen
Mandiant, “UNC5537 Targets Snowflake Customer Instances for Data Theft and Extortion,” Google Cloud Blog, 11 June 2024. Mandiant’s finding is for the campaign against Snowflake customers as a whole; it does not name individual victims.
- 7
listed for sale
Waqas, “Ticketmaster Data Breach: Hackers Selling 560 Million Users Data for $500,000,” Hackread, 28 May 2024; CBS News/AFP, “Hacking Group Claims It Breached Ticketmaster and Stole Data for 560 Million Customers,” CBS News, updated 30 May 2024.
hackread.com/hackers-ticketmaster-data… ↗cbsnews.com/news/ticketmaster-breach-… ↗
- 8
breaches that took more than 200 days
IBM Security, Cost of a Data Breach Report 2025 (Armonk, NY: IBM, 2025). US$5.01 million against US$3.87 million.
- 9
One organisation in five has had a breach linked to shadow AI
IBM Security, Cost of a Data Breach Report 2025 (Armonk, NY: IBM, 2025). The US$670,000 compares organisations with high levels of shadow AI with those with low levels or none; IBM puts shadow AI’s addition to the global average at US$200,000.
- 10
twenty-two seconds
Jurgen Kutscher, “M-Trends 2026: Data, Insights, and Strategies From the Frontlines,” Google Cloud Blog, 23 March 2026. Mandiant measures the time between an initial access event and the hand-off to a secondary threat group, and calls the first group initial access partners.
- 11
under ten days
Anamarija Pogorelec, “Attackers Are Handing Off Access in 22 Seconds, Mandiant Finds,” Help Net Security, 24 March 2026. As reported by Help Net Security from the full M-Trends 2026 report.
- 12
Cybersecurity Ventures
Taylor Fox, “35,000 Chief Information Security Officers Employed Globally in 2026,” Cybersecurity Ventures, 23 March 2026 (announcing the 2026 CISO Report from Cybersecurity Ventures in partnership with Sophos). Produced in partnership with Sophos, a security vendor; the report positions MSPs/MSSPs as the answer for SMBs.
- 13
CISO-Board Engagement Report
IANS Research, “New Report Reveals Key Gaps in Board-CISO Strategic Dialogue on Cyber Risks,” press release, Boston, 3 March 2026; IANS Research, “Boards Give CISO Cybersecurity Reporting a Mixed Grade,” IANS blog, 24 March 2026. IANS sells research and advisory services to CISOs; Artico Search recruits security executives; The CAP Group advises on cyber and board matters.
iansresearch.com/resources/press-releases/… ↗iansresearch.com/resources/all-blogs/post/… ↗
- 14
Splunk
Splunk, “Splunk Report: CISOs Gain Influence in the C-Suite and Boardrooms Worldwide,” press release, San Francisco, 23 January 2025 (updated 21 February 2025); Cisco Systems, Inc., “Splunk Report: CISOs Gain Influence in the C-Suite and Boardrooms Worldwide,” PR Newswire, 23 January 2025. Splunk (a Cisco company) sells security and observability software to CISOs.
splunk.com/en_us/newsroom/press-rele… ↗prnewswire.com/news-releases/splunk-repo… ↗
- 15
McKinsey
McKinsey & Company, “State of AI Trust in 2026: Shifting to the Agentic Era,” Tech Forward (blog), 25 March 2026.
- 16
AI Risk Management Framework
National Institute of Standards and Technology, Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1 (Gaithersburg, MD: NIST, January 2023).
- 17
ISO 42001 is the standard you can be certified against
International Organization for Standardization, ISO/IEC 42001:2023, Information Technology – Artificial Intelligence – Management System, ed. 1 (Geneva: ISO, December 2023); International Organization for Standardization, “ISO/IEC 42001 Explained,” ISO.
iso.org/standard/42001 ↗iso.org/home/insights-news/resour… ↗
- 18
The EU AI Act is the legal floor
Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act), Official Journal of the European Union L, 12 July 2024, arts. 2(1)(c), 26, 99.
- 19
The OECD principles
OECD, Recommendation of the Council on Artificial Intelligence, OECD/LEGAL/0449, adopted 22 May 2019, revised 3 May 2024, OECD Legal Instruments.
- 20
10 December 2026
Privacy and Other Legislation Amendment Act 2024 (Cth), No. 128, 2024, sch. 1, pt. 15 (inserting Australian Privacy Principles 1.7–1.9 into the Privacy Act 1988).
- 21
CPS 230 and CPS 234
Australian Prudential Regulation Authority, “Operational Risk Management,” APRA; Australian Prudential Regulation Authority, Prudential Standard CPS 230 Operational Risk Management, in force 1 July 2026, APRA Handbook; Australian Prudential Regulation Authority, Prudential Standard CPS 234 Information Security, in force 1 July 2019, APRA Handbook. CPS 230 commenced on 1 July 2025; the current version commenced on 1 July 2026.
apra.gov.au/operational-risk-management ↗handbook.apra.gov.au/standard/cps-230 ↗handbook.apra.gov.au/standard/cps-234 ↗
- 22
HiddenLayer
HiddenLayer, “HiddenLayer Releases the 2026 AI Threat Landscape Report, Spotlighting the Rise of Agentic AI and the Expanding Attack Surface of Autonomous Systems,” press release, Austin, TX, 18 March 2026. HiddenLayer’s category is public model and code repositories; the 35 per cent is the share of respondents citing it. HiddenLayer sells security for AI models, including model-supply-chain scanning.
- 23
IBM found 13 per cent
IBM Security, Cost of a Data Breach Report 2025 (Armonk, NY: IBM, 2025).
- 24
Nutanix’s 2026 survey of the sector
Nutanix, “AI Adoption in Financial Services Is Accelerating, but Governance Gaps and Infrastructure Challenges Are Slowing Scale: Nutanix Enterprise Cloud Index,” press release, 2026, reprinted in Financial IT. The release’s wording: 66 per cent of IT executives report employees using unsanctioned AI. Nutanix sells hybrid-cloud infrastructure.
- 25
German SMEs
Wolters Kluwer Tax & Accounting, “Wolters Kluwer Research: Germany’s SMEs Put Security before Speed in Digital Transformation,” press release, Ludwigsburg, Germany, Business Wire, 24 March 2026, reprinted in The AI Journal. The release reports the figures; the reasons behind them are not part of the survey. Wolters Kluwer sells tax, accounting and compliance software and services to SMEs and their advisers.
Chapter 6Becoming the Future-Ready Leader6 notes
- 1
Kim at the London School of Economics
Hye-young Kim and Ann L. McGill, “AI-Induced Dehumanization,” Journal of Consumer Psychology 35, no. 3 (2025): 363–81. In the study participants watched a short video of a robot; the treatment they rated was of employees in a workplace scenario.
- 2
In a global Deloitte survey
Deloitte Global, “Boards and C-suite Cite Open Communication, Risk Management, and Scenario Planning as Critical to Building Organizational Resilience,” press release announcing How Board and C-suite Collaboration Can Build Organizational Resilience, Deloitte Global Boardroom Program with the Deloitte Global CEO Program, 30 September 2025. 739 board members and C-suite executives in more than 50 countries; 71 per cent said this is where board oversight can help boost resilience the most.
- 3
State of Data and AI Literacy Report
DataCamp, “Companies Are Investing in AI, But Their Workforces Aren’t Ready, According to New DataCamp/YouGov Report,” press release, Business Wire, 26 February 2026; DataCamp, “Data & AI Literacy in 2026: Stats and Skills Gap,” DataCamp blog, 2026. YouGov survey of 517 US and UK business leaders; the 21 per cent is the whole sample. DataCamp sells data and AI literacy training.
businesswire.com/news/home/20260226726072/… ↗datacamp.com/blog/the-state-of-data-an… ↗
- 4
Bill Briggs
Nick Lichtenberg, “Deloitte’s CTO on a Stunning AI Transformation Stat: Companies Are Spending 93% on Tech and Only 7% on People,” Fortune, 15 December 2025; Deloitte Insights, Tech Trends 2026 (Deloitte, 2025). Deloitte attributes the split to its own research on AI investment allocation, 2025, which is not published.
fortune.com/2025/12/15/deloitte-cto-b… ↗mkto.deloitte.com/rs/712-CNF-326/images/DI_… ↗
- 5
10-20-70
Boston Consulting Group, “The Leader’s Guide to Transforming with AI,” 12 December 2024.
- 6
Future of Jobs Report 2025
World Economic Forum, Future of Jobs Report 2025, Insight Report (Geneva: World Economic Forum, January 2025).
Chapter 7Leading the Transformation19 notes
- 1
EY polled
Ernst & Young LLP, “EY survey: autonomous AI adoption surges at tech companies as oversight falls behind,” press release, EY US, 4 March 2026. EY reports the 52 per cent of department-level initiatives without formal approval and the 45 per cent reporting a confirmed or suspected leak as separate findings.
- 2
State of Organizations
McKinsey & Company, The State of Organizations 2026 (McKinsey & Company, February 2026), 7.
- 3
automation bias
Kathleen L. Mosier, Linda J. Skitka, Susan Heers, and Mark Burdick, “Automation Bias: Decision Making and Performance in High-Tech Cockpits,” International Journal of Aviation Psychology 8, no. 1 (1998): 47–63.
- 4
learned carelessness
Frederik Aust, Christoph Moehlenbrink, and Meike Jipp, “Operationalization of Learned Carelessness,” Proceedings of the Human Factors and Ergonomics Society Annual Meeting 55, no. 1 (2011): 1735–39.
- 5
S&P Global
S&P Global Market Intelligence, “Generative AI experiences rapid adoption, but with mixed outcomes – Highlights from VotE: AI & Machine Learning,” Research, 30 May 2025.
- 6
four out of five AI projects
James Ryseff, Brandon F. De Bruhl, and Sydne J. Newberry, The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed: Avoiding the Anti-Patterns of AI, RR-A2680-1 (Santa Monica, CA: RAND Corporation, 2024).
- 7
at Meta in March 2026
Stevie Bonifield, “A rogue AI led to a serious security incident at Meta,” The Verge, 19 March 2026; Amanda Silberling, “Meta is having trouble with rogue AI agents,” TechCrunch, 18 March 2026. Both reports give about two hours as the time the data was exposed.
theverge.com/ai-artificial-intelligenc… ↗techcrunch.com/2026/03/18/meta-is-having… ↗
- 8
Gravitee
Gravitee, The State of AI Agent Security 2026 (Gravitee, 2026); Jorge Ruiz, “State of AI Agent Security 2026 Report: When Adoption Outpaces Control,” Gravitee (blog), 4 February 2026. The report (919 respondents) gives 24.4 per cent with full visibility of which agents interact with each other, 21.9 per cent treating agents as identity-bearing entities, and 3.9 per cent monitoring and securing more than 80 per cent of their agents. Gravitee sells API and AI-agent management and security gateway tooling.
gravitee.io/hubfs/Downloadable Resour… ↗gravitee.io/blog/state-of-ai-agent-se… ↗
- 9
SailPoint, which sells
SailPoint, “SailPoint research highlights rapid AI agent adoption, driving urgent need for evolved security,” press release, 28 May 2025. Survey of 353 participants with enterprise security responsibilities; 80 per cent of companies said their AI agents had taken unintended actions. SailPoint sells identity security, including governance of AI agent identities.
- 10
published a deployment blueprint
Microsoft, “Secure & Governed Data Foundation for Microsoft Copilot - Foundational Deployment Guidance,” Microsoft Learn, last updated 18 August 2026; Microsoft, “Get ready for Microsoft Copilot and agents with SharePoint Advanced Management,” Microsoft Learn, last updated 16 July 2026. The blueprint has three pillars: remediate oversharing, set up guardrails, meet regulations.
learn.microsoft.com/en-us/microsoft-365/copil… ↗learn.microsoft.com/en-us/sharepoint/get-read… ↗
- 11
Concentric AI
Concentric AI, “Latest Industry Data Risk Report from Concentric AI Shows 60% Increase in Oversharing of Sensitive Data over the Past Year,” press release, 23 February 2023; Concentric AI, Concentric AI Data Risk Report 2H 2025 (Concentric AI, 2025). The 16 per cent and 802,000 at-risk files are from Concentric’s 2022 report; the Copilot, 57 per cent and 70 per cent figures are from its 2H 2025 report, and both describe Concentric’s own customers. Concentric AI sells data security posture management, including Microsoft Copilot data governance.
concentric.ai/press-release/latest-indu… ↗concentric.ai/wp-content/uploads/2024/0… ↗
- 12
ISACA
ISACA, “Adopted, not governed: new ISACA research reveals AI blind spot at the heart of enterprise risk,” press release, 23 March 2026.
- 13
SANS and GIAC
SANS Institute and GIAC Certifications, 2026 Cybersecurity Workforce Research Report by SANS | GIAC (SANS Institute, 11 March 2026); SANS Institute, “SANS Research: The Cybersecurity Talent Shortage Narrative Is Wrong. The Real Crisis Is What Your Team Doesn’t Know, Starting with AI,” press announcement, 2026. Skills gaps led headcount 60 to 40 in 2026, up from 52 to 48 in 2025. Respondents were security leaders (72 per cent), HR and talent professionals (16 per cent) and people with both roles (12 per cent). SOC and security analyst roles lead AI-driven role reductions at 32 per cent. SANS and GIAC sell cybersecurity training and certification, the remedy for the skills gap the report finds.
sans.org/white-papers/2026-cyberse… ↗sans.org/press/announcements/sans-… ↗
- 14
app builder collapsed
Amanda Silberling, “Once worth over $1B, Microsoft-backed Builder.ai is running out of money,” TechCrunch, 20 May 2025; Codekeeper, “Builder.ai collapse: Why you need AI escrow in 2026”. The insolvency is reported by TechCrunch; the account of customers losing access to their software comes from Codekeeper. Codekeeper sells software escrow, the remedy for the risk it describes.
techcrunch.com/2025/05/20/once-worth-ove… ↗codekeeper.co/articles/builderai-collap… ↗
- 15
agentic supply chain vulnerabilities
OWASP Gen AI Security Project, OWASP Top 10 for Agentic Applications 2026, version 2026 (OWASP Foundation, December 2025), ASI04.
- 16
data colonialism
Nick Couldry and Ulises A. Mejias, “Data Colonialism: Rethinking Big Data’s Relation to the Contemporary Subject,” Television & New Media 20, no. 4 (2019): 336–49; see also Michael Kwet, “Digital Colonialism: US Empire and the New Imperialism in the Global South,” Race & Class 60, no. 4 (2019): 3–26. Couldry and Mejias’s term is broader than the country-level sense used here; Kwet’s “digital colonialism” is specifically about the Global South.
doi.org/10.1177/1527476418796632 ↗doi.org/10.1177/0306396818823172 ↗
- 17
HVAC
U.S. Senate Committee on Commerce, Science, and Transportation, A “Kill Chain” Analysis of the 2013 Target Data Breach, Majority Staff Report for Chairman Rockefeller, 26 March 2014.
- 18
Talos
Aliza Johnson and James Nutland, “Talos IR ransomware engagements and the significance of timeliness in incident response,” Cisco Talos Blog, 16 July 2025. In the first engagement data was stolen but encryption was prevented; in the second the actors achieved nearly 100 per cent host encryption.
- 19
Sygnia
Sygnia, “73% of CISOs Unprepared for the Next Big Cyber Attack, Incident Response Readiness Report Reveals,” press release, 13 April 2026; Rena Stern, “Why 73% of Organizations Still Aren’t Ready for a Cyberattack,” Sygnia (blog), 28 June 2026. The survey question was whether the organisation would be fully ready to execute its response if a significant attack occurred tomorrow. Sygnia sells incident response and readiness services.
sygnia.co/press-release/sygnia-rele… ↗sygnia.co/blog/why-organizations-st… ↗
Chapter 8Your Role in Shaping What Comes Next14 notes
- 1
roughly 665 million people
Global Entrepreneurship Monitor, GEM 2024/2025 Global Report: Entrepreneurship Reality Check (London: GEM, 2025). GEM reports a rate, not a headcount: roughly one in eight working-age adults was starting or running a business in 2024. The 665 million figure is the widely quoted estimate that follows from that rate.
- 2
reward hacking
Victoria Krakovna et al., “Specification gaming: the flip side of AI ingenuity,” Google DeepMind (blog), 21 April 2020. The homework example is DeepMind’s.
- 3
84 per cent
Stack Overflow, “AI,” 2025 Stack Overflow Developer Survey, 29 July 2025; Stack Overflow, “Developers remain willing but reluctant to use AI: The 2025 Developer Survey results are here,” Stack Overflow Blog, 29 December 2025 (republished).
survey.stackoverflow.co/2025/ai ↗stackoverflow.blog/2025/12/29/developers-rem… ↗
- 4
roughly 45 per cent of AI-generated code
Veracode, “Insights from 2025 GenAI Code Security Report,” Veracode Blog, 30 July 2025; full report: Veracode, 2025 GenAI Code Security Report (October 2025 update); Veracode, “Spring 2026 GenAI Code Security Update: Despite Claims, AI Models Are Still Failing Security,” Veracode Blog, 24 March 2026. The spring 2026 update reports pass rates flat at 45 to 55 per cent across model generations, with GPT-5 reasoning models the exception at 70 to 72 per cent. Veracode sells application-security testing.
veracode.com/blog/genai-code-security-… ↗veracode.com/resources/analyst-reports… ↗veracode.com/blog/spring-2026-genai-co… ↗
- 5
Cyble
Cyble Research and Intelligence Labs, “When AI Secrets Go Public: The Rising Risk of Exposed ChatGPT API Keys,” Cyble (blog), 12 February 2026. Cyble says such sites “frequently” call the API directly from the browser. Cyble sells threat intelligence and external exposure monitoring.
- 6
Fast-fashion operators
Alina Selyukh, “America can’t resist fast fashion. Shein, with all its issues, is tailored for it,” NPR, 13 October 2023. NPR reports Shein adding up to 10,000 new items a day and turning a design into a garment in as little as ten days.
- 7
ONDC
Indian Retailer Bureau, “ONDC Reports 10 Mn Monthly Transactions in June,” Indian Retailer, 9 July 2024; Akshita Toshniwal, “Monthly transactions on ONDC to hit around 50 million by the end of the year: CEO Koshy,” YourStory, 29 May 2024; Press Information Bureau, “2024 Year End Review for Department for Promotion of Industry and Internal Trade,” Ministry of Commerce & Industry, Government of India, 20 December 2024.
indianretailer.com/news/ondc-reports-10-mn-m… ↗yourstory.com/2024/05/50-million-monthl… ↗pib.gov.in/PressReleasePage.aspx?PRI… ↗
- 8
Manufacturers building machine learning
Protolabs, Innovation in Manufacturing 2026: How manufacturing’s digital adoption is transforming the product life cycle (Protolabs, 2026). Protolabs describes its report as a review of existing reports and industry data. Protolabs sells digital/on-demand manufacturing services.
- 9
Global AI Confessions Report
Dataiku, “78% of CEOs Say AI Could Cost Them Their Job and Their Company’s Future, Finds Dataiku Global AI Confessions Report,” press release, Business Wire, 4 May 2026. The 80 per cent is the release’s figure for CEOs who say their job will be at risk by the end of 2026; the 78 per cent in its title is a separate measure. Dataiku sells an enterprise AI platform; the survey was commissioned by Dataiku.
- 10
Samsung engineers
Laura Dobberstein, “Samsung reportedly leaked its own secrets through ChatGPT,” The Register, 6 April 2023; Arjun Kharpal, “Samsung bans use of A.I. like ChatGPT for employees after misuse of the chatbot,” CNBC, 2 May 2023. CNBC reports a temporary restriction on generative AI on company devices, communicated by memo to one of Samsung’s largest divisions.
theregister.com/2023/04/06/samsung_report… ↗cnbc.com/2023/05/02/samsung-bans-u… ↗
- 11
UpGuard, which sells tools
UpGuard, The State of Shadow AI (UpGuard, 10 November 2025); press release: “New Research from UpGuard Reveals 68% of Security Leaders Admit to Unauthorized AI Usage,” 10 November 2025. The survey does not measure organisations that banned AI. UpGuard sells cyber risk and AI-usage governance tooling.
upguard.com/resources/the-state-of-sh… ↗upguard.com/press/new-research-from-u… ↗
- 12
On 21 June 2026
OpenAI, “Samsung Electronics brings ChatGPT and Codex to employees,” OpenAI, 21 June 2026; PYMNTS, “Samsung Rolls Out OpenAI Tools to Workforce in Major Enterprise Push,” PYMNTS.com, 22 June 2026. PYMNTS gives about 125,000 as the employees in South Korea, with the Device eXperience division’s international staff in addition.
openai.com/index/samsung-electronics… ↗pymnts.com/news/artificial-intellige… ↗
- 13
NIS2
Directive (EU) 2022/2555 of the European Parliament and of the Council of 14 December 2022 on measures for a high common level of cybersecurity across the Union (NIS 2 Directive), OJ L 333, 27.12.2022, 80–152, art. 20(1).
- 14
DORA
Regulation (EU) 2022/2554 of the European Parliament and of the Council of 14 December 2022 on digital operational resilience for the financial sector (DORA), OJ L 333, 27.12.2022, 1–79, arts. 5 and 28.