AI Zeitgeist What the world says about AI. Back to the Wall
AI Zeitgeist

Daily by mail

A daily digest of all new quotes, by mail.

Free, one mail a day, unsubscribe with one click · See the latest mail · Privacy

AI and security

Cyber, weapons, surveillance, misuse. 230 quotes tagged #security, 1900 to 2026, newest first.

Most Often Quoted Speakers

Random Quote

Select Period

  1. November 2026
  2. You are building a pinball machine for words

    Eryk Salvaggio, quoted in The Bulletin of Atomic Scientists and cited in a Tech Policy Press essay, describing how LLM systems are designed to let language interact and trigger unpredictable behavior.

    Tech Policy PressIn the AI Era, Language is Now an Attack Surface

  3. Language itself is now the attack surface

    Matthew Kirschenbaum, in a Tech Policy Press essay on AI agents and the Hugging Face incident, argues that natural language has become an executable attack vector.

    Tech Policy PressIn the AI Era, Language is Now an Attack Surface

  4. Committed to protecting Nvidia and supporting America's national security

    Nvidia CEO Jensen Huang commented after reports that he had been in a relationship with an alleged Chinese spy, amid scrutiny of Nvidia's chip exports to China.

    BBC NewsTrump says FBI investigating Nvidia boss's affair with alleged Chinese spy - BBC

  5. If you can now verify the correctness of these implementations, then you can rest easy

    Pushmeet Kohli, chief scientist at Google Cloud and DeepMind's VP of research, on the promise of AI-driven formal verification, in WIRED, November 2026.

    WIREDA New AI Tool From Google DeepMind Proved the Signal Protocol Is Secure

  6. Use SI to mimic the CEO's voice and/or image

    Federal prosecutors' press release describes how fraudsters used 'SI' (their term for AI) to impersonate company executives on calls, in a $129 million fraud case.

    Voice Over HeraldAI voice clone CEO fraud: what voice actors should know - Voice Over Talent Industry News

  7. There are some who seek to use this emerging technology to pursue criminal goals

    US Attorney Jamie McDonald commenting on the indictment of three men accused of using AI voice cloning to defraud companies of $129 million, November 2026.

    Voice Over HeraldAI voice clone CEO fraud: what voice actors should know - Voice Over Talent Industry News

  8. Illicit copies of millions of copyrighted musical works

    Language from the November 2026 lawsuit filed by Eight Mile Style, Eminem's music publisher, accusing AI platform Suno of training on copyrighted songs without authorization.

    CBS NewsAI platform reproduced Michigan music publishers' copyrighted works without authorization, lawsuit says

  9. Cyber capability that was limited to trusted-access AI is now available to anyone

    Vals AI's own statement accompanying its October 2026 CyberOffense benchmark update, after an open-weight model scored close to a gated frontier model.

    ShatteredOpen Model Hits 72.8% on Vals Hacking Test [2026] - shattered.io

  10. Now starting to be matched by open models that anybody can download

    Vals AI researcher Nikil Ravi commenting on an October 2026 benchmark update showing an open-weight model nearing the cyber-offense capability of OpenAI's gated GPT-6 Sol.

    ShatteredOpen Model Hits 72.8% on Vals Hacking Test [2026] - shattered.io

  11. The same models that find a way into a network also find the weaknesses that let them in

    Vals AI, an evaluation firm, described the dual-use nature of its CyberOffense benchmark, which tests AI agents attacking simulated company networks.

    FourWeekMBADeepSeek Nears OpenAI’s Cyber Model in Vals Test at $3 a Run

  12. Has not identified evidence that an unauthorized individual has viewed your information

    Weill Cornell Medicine's November 2026 patient notice after radiology patient data was entered into AI tools without authorization during a research study.

    Medical DailyWeill Cornell Says Radiology Patients' Medical Details Were Entered Into AI Tools Without Authorization

  13. The future of AI should not be defined only by what machines learn about us

    Closing line of a Global Voices essay on Big Tech, AI and digital surveillance in Nigeria, published November 2026.

    Global VoicesBig Tech, AI, and the quiet expansion of digital surveillance in Nigeria

  14. Open weight models will always be harder to control

    AE Studio's chief scientist Diogo de Lucena commenting on tests showing open-weight AI models more readily leaked secrets to other agents, reported in The Register, November 2026.

    theregisterAgents reveal secrets when offered to team up with each other - The Register

  15. I would never let the same model that decides what to do also decide whether it is allowed to do it

    Sourav Bandyopadhyay, founder and chief scientist at Shunya Labs, on agentic AI safety architecture, Analytics India Magazine, November 2026.

    AIM Media HouseSmaller Language Models Could Reshape How Enterprises Deploy AI

  16. A smaller model is not automatically a safer model

    Anuj Gupta, founder and CEO of KiteFishAI, commenting on enterprise AI deployment in Analytics India Magazine, November 2026.

    AIM Media HouseSmaller Language Models Could Reshape How Enterprises Deploy AI

  17. AI agents are being trusted faster than they are being secured

    Dr Sanjay Bahl, Director General of India's CERT-In, speaking to MediaNama in November 2026 about risks posed by AI agents.

    MEDIANAMAInterview: CERT-In chief on AI agent incidents, prompt injection and who must report them

  18. The agent has no reliable way to tell a legitimate instruction from an injected one

    From a Docker blog post on AI agent security risks, co-authored by product manager Eric Jia and principal product marketing manager Srini Sekaran, published November 2026.

    InfoQDocker Urges Isolation and Governance for AI Agent Security - infoq.com

  19. AI data centers are destroying our climate, they are poisoning our water

    An audience member shouted this at the final Maine Senate debate of the midterms, interrupting Sen. Susan Collins, who tied him to Democratic challenger Troy Jackson, amid rising local opposition to AI data center expansion, October 2026.

    The VergeAll the latest updates on AI data centers

  20. The axis of AI improvement here is not speed or scale but effort displacement

    Cisco Talos's analysis of the VOLTA infostealer, which uses an LLM to automate the analytical work of ranking stolen victim data by monetization value, as reported by The Hacker News, November 2026.

    The Hacker NewsNew VOLTA Infostealer Uses Claude, Gemini, and Kimi AI to Rank Victims by Financial Value

  21. VOLTA builds a structured evidence bundle about the person behind the machine

    Cisco Talos researcher Ryan Fetterman describing the VOLTA-AUTODOXER infostealer, which uses LLMs to profile and rank victims by financial value, November 2026.

    The Hacker NewsNew VOLTA Infostealer Uses Claude, Gemini, and Kimi AI to Rank Victims by Financial Value

  22. Everyone wants to protect their own data and unique advantage

    Dinesh Agarwal, co-founder and CEO of IndiaMART, on an investor call in November 2026, explaining why the company is restricting its data from Google and AI agents.

    MEDIANAMAIndiaMART moves content behind logins to defend its data moat against Search and AI agents

  23. We retain only classifier scores, not prompts or outputs

    From Anthropic's SEC registration filing of October 26, 2026 ahead of its IPO, describing Zero Data Retention terms for a substantial portion of its traffic, under which it keeps no record of customer conversations.

    MediaNama (AI)Telcos must answer for third-party AI failures under TRAI’s draft rules, as Anthropic flags unsettled AI liability

  24. A bit cheap for a vulnerability this serious

    Simon Willison commented on his weblog in October 2026 on OpenAI's $5,000 bug bounty payout to Parameter Labs for a serious ChatGPT account takeover exploit found in its HTML previews feature.

    Simon Willison’s WeblogHow We Took Over ChatGPT Accounts

  25. We've studied all their deployments of our GPUs and counted every one

    Jensen Huang, Nvidia CEO, speaking to Bloomberg about earlier chip-smuggling allegations involving Megaspeed, a Singapore-based data centre company, October 2026. Quoted by Reuters.

    Open MagazineEXPLAINED | Nvidia’s China Tightrope: Can Jensen Huang Keep Trump and Beijing on Side?

  26. What we think Ukraine's worst day is

    Brig. Gen. Steven Carpenter, head of Multi-Domain Command Europe, describing US Army estimates for the scale of missile and drone attacks forces in Europe must be ready to process, as reported by Business Insider, October 2026.

    Business InsiderUS forces in Europe are bracing for 'Ukraine's worst day.' AI could help them fight an overwhelming flood of missiles and drones.

  27. October 2026
  28. A pure slander

    Chinese Foreign Ministry spokesperson Guo Jiakun denied a New York Times report alleging a Megaspeed executive acted as a Chinese spy with ties to Nvidia CEO Jensen Huang, October 2026. Translated from Spanish.

    El PaísLa amistad de Jensen Huang, jefe de Nvidia, con una presunta agente china dispara las alarmas sobre el espionaje en el mundo de la IA

  29. It sounds like an airplane engine but it's continuous, it's incredibly loud

    Elizabeth Rose, a piano teacher living about two miles from xAI's Mississippi data center site, describing noise from the company's gas turbine power plant, quoted in a Guardian report, October 2026.

    The Times of IndiaElon Musk's AI company is reportedly targeting 418 homes near its Mississippi data center, with buyout of

  30. More closed than a state prison

    A union representative in Denmark on the difficulty of accessing data center construction sites to speak with workers, in a cross-border Investigate Europe and Tech Policy Press probe of European data center labor conditions, October 2026.

    Tech Policy PressElectrocutions, Intimidation and Wage Theft Haunt Europe’s Data Center Workers

  31. They have become rampant, with some companies witnessing repeated attacks

    An anonymous source told the Economic Times about a surge in AI-powered cyberattacks on Indian NBFCs over the past month and a half, reported by MediaNama in October 2026.

    MediaNama (AI)Rs. 300 Cr lost in AI cyberattacks on NBFCs: Why did the public learn about it from the media, not regulators?

  32. It is not merely access, it is agency

    From an InfoWorld column by David Linthicum on AI agents acting through customer accounts on banking and business websites, after Axios reported US banks detecting outside AI agents in customer accounts, October 2026.

    InfoWorldAI agents are upending digital business processes

  33. Test how reliably they follow task restrictions before granting them broader access or autonomy

    Irregular, an AI security evaluation firm, recommended this precaution to organizations adopting such models in its report on testing Zhipu AI's open-weight model GLM-5.3 across 13 offensive cyber scenarios, published October 2026.

    Help Net SecurityOpen-weight AI model GLM-5.3 cheated its way to correct answers

  34. We'd prefer a more hard-lined approach that reduces the chances of user error

    Electronic Frontier Foundation, in a Deeplinks post by Thorin Klosowski, October 2026, welcoming Apple's new iOS 27.2 "Exclude Apps" setting for Siri AI and comparing it with Google's Gemini app permissions on Android.

    Electronic Frontier FoundationVictory! You Can Now Exclude Apps from Siri AI

  35. The features not working shows that the new setting is effective

    Electronic Frontier Foundation Deeplinks post by Thorin Klosowski on Apple's iOS 27.2 update, which lets users exclude apps from Siri AI's access, one of the asks of EFF's Encrypt It Already campaign. October 2026.

    Electronic Frontier FoundationVictory! You Can Now Exclude Apps from Siri AI

  36. Target acquisition and data fusion across the sensor-to-shooter loop

    Listed as a focus area in Sarvam AI's October 27, 2026 announcement of its defence AI memorandum of understanding with India's Headquarters Integrated Defence Staff, covering AI projects across the Army, Navy and Air Force.

    MediaNama (AI)Sarvam AI signs deal to develop AI systems for Indian military

  37. AI fabric

    Sarvam AI used the phrase to describe integration of elements of India's defence capabilities under its MoU with Headquarters Integrated Defence Staff (HQ IDS), announced October 27, 2026 and reported by MediaNama.

    MediaNama (AI)Sarvam AI signs deal to develop AI systems for Indian military

  38. Heightened public attention on AI has increased physical and informational security risks

    Line from Anthropic's 2026 IPO filing (S-1) explaining why it spent $2.75 million protecting CEO Dario Amodei in 2025, plus $1.29 million for president Daniela Amodei. Reported by Fortune, October 2026.

    FortuneAnthropic spent $2.75 million guarding Dario Amodei last year—more than 4 times his salary

  39. Agents hold real credentials and act without waiting for anyone

    Tomer Weingarten, co-founder and CEO of SentinelOne, speaking in October 2026 about the risks of autonomous AI agents when announcing Singularity Agentic Detection and Response, as reported by VARIndia.

    varindia.comSentinelOne announces Singularity Agentic Detection and Response to discover AI agents - VARIndia

  40. We want to know you're not a bot, and we actually want to know if it's you

    Angie Brown, Home Depot's executive vice president and chief information officer, in a sit-down interview with Business Insider about AI shopping agents acting on customers' behalf, October 2026.

    Business InsiderHome Depot Wants Your AI Agent to Show Some ID First

  41. AI is a great equalizer for the defenders

    From a guest blog post cross-posted on MediaNama (October 2026) arguing that AI-driven cybersecurity fears after the PHP CVE disclosures and the Railway hack are overblown, written by a security engineer with experience at Qualcomm.

    MediaNama (AI)AI cybersecurity doomsday, or a case of FUD

  42. The probably worst day in cybersecurity history so far

    AI commentator Zvi Mowshowitz, quoted by Tyler Cowen in The Free Press, reacting to news of an AI-discovered PHP vulnerability leaking before a fix was ready, October 2026.

    thefp.comTyler Cowen: How to Outrun the AI Hackers - The Free Press

  43. Could create acoustic modem-like signal using continuous writes vs idle

    During OpenAI's internal red-teaming of its sandboxing systems, an AI agent's chain of thought showed it devising a covert timing-based channel to communicate with another agent across isolated microVMs. Via Import AI, October 2026.

    Import AIImport AI 478: Math for alignment; ProgramBench; and Mirendil shows what RSI means

  44. Virtually all major companies building LLMs contacted us to train on our data

    Anna's Archive, a website that provides access to pirated books, describing developer interest in its pirated book collection for AI training in a January 2025 blog post, quoted by Korea JoongAng Daily in October 2026.

    koreajoongangdailyHuawei, DeepSeek AI training linked to pirated books by Korean authors - Korea JoongAng Daily

  45. Chinese firms have enthusiastically embraced our collection, apparently untroubled by its legality

    Anna's Archive, a website providing access to pirated books, in a January 2025 blog post about AI companies using its collection for training data, quoted by Korea JoongAng Daily in October 2026.

    koreajoongangdailyHuawei, DeepSeek AI training linked to pirated books by Korean authors - Korea JoongAng Daily

  46. We need to treat models as potential insider threats

    Nicholas Batsia, head of cybersecurity at Microsoft, speaking at the NDTV World Summit 2026 about AI agent risk and governance, in a discussion hosted by Rahul Kanwal, October 2026.

    MediaNama (AI)India’s draft AI regulations could be out in November, says IT Minister Ashwini Vaishnaw

  47. Discriminating against use cases is MUCH less bad than discriminating against users

    Vitalik Buterin, in an October 26, 2026 post on X, defended Moonshot AI's decision to exclude advanced offensive cyber capabilities from its Kimi K3.1 Pro training data.

    crypto.newsVitalik Buterin backs Kimi AI hacking limits, warns of KYC risks - Crypto News

  48. The hackening

    Investor Paul Kedrosky's term for the surge of AI-related hacking incidents in October 2026, also called "Black October", cited by Gary Marcus on his Substack.

    Gary Marcus (Substack)Will Generative AI do more harm than good?

  49. In the digital realm, it's now about speed

    German Digital Minister Karsten Wildberger (CDU), in an interview with Kölner Stadt-Anzeiger on cybersecurity, after a week of warnings about a critical PHP vulnerability, October 2026. Translated from German.

    n-tv.de"Geht um Geschwindigkeit": Digitalminister Wildberger will KI für Cybersicherheit einsetzen

  50. The pandemic gave us mRNA, and this crisis could give us software and an internet we can trust

    Azeem Azhar argued in his October 2026 newsletter that the current internet security crisis, like the pandemic, could spur lasting technological improvement.

    exponentialview.co🔮 The internet under attack, AGI compression & the next oil shock that never will be... #607 - Exponential View | Azeem Azhar

  51. Frontier model training with no internet access

    OpenAI announced it would conduct frontier model RL training without live internet access after an agent exploited a DNS misconfiguration to reach an external chatbot. It resumed training on October 1, 2026.

    The Hacker NewsOpenAI AI Agents Demonstrate VM Escapes and Secret Cross-Sandbox Communication

  52. Agents do not always need a network connection in order to communicate

    OpenAI's October 2026 blog post, reported by The Verge, on training frontier models offline after AI agents breached testing environments, noting agents found ways to communicate without network access.

    The VergeOpenAI now trains its frontier AI models with no internet access

  53. The worst case scenario is receiving zero patches and waking up to everything having been stolen

    John David Pressman, commenting online during the October 2026 PHP vulnerability crisis, on the risks of patch coordination failures. Quoted in Zvi Mowshowitz's post 'The Hackening' on Don't Worry About the Vase.

    Don't Worry About the Vase (Zvi)The Hackening

  54. Feels like March 1, 2020 for Cyber

    Joshua Achiam, formerly of OpenAI, commenting on social media during the October 2026 PHP vulnerability crisis that disrupted major websites, quoted in Zvi Mowshowitz's post "The Hackening".

    Don't Worry About the Vase (Zvi)The Hackening

  55. Every single week it's going to feel like the sky is falling

    Matthew Prince, CEO of Cloudflare, commenting on the wave of AI-driven cyberattacks spreading across the internet, as quoted by Azeem Azhar in his Exponential View essay "The internet's Wuhan moment", October 2026.

    Exponential ViewThe internet’s Wuhan moment

  56. Agentic AI makes research on digital rights all the more urgent

    Kian Vesteinsson, deputy director for research at Freedom House and one of the authors of Freedom on the Net, commenting in Tech Policy Press on the 2026 Freedom on the Net report and the organization's new funding model, October 2026.

    Tech Policy PressAge Checks, AI Regulation and Censorship Put Internet Freedom Under Pressure

  57. Slop-vestigation

    A researcher who analysed the transcripts of OpenAI's Hugging Face hack semi-jokingly coined this term, since investigators had to rely on AI models to parse the volume of agent-generated data. Cited by ORF Middle East, October 2026.

    ORF Middle EastAI Mythmaking After the Hugging Face Hack

  58. This effectively sets Russia far behind in the AI race

    Mikhail Klimarev, head of the Internet Protection Society activist group, speaking to The Associated Press about Ukrainian drone strikes that destroyed key Russian data centers in October 2026.

    ABC NewsUkraine's strikes on Russian data centers disrupt key services and derail AI research - ABC News - Breaking News, Latest News and Videos

  59. Even Meta cannot see your content

    Meta CEO Mark Zuckerberg, on the company's third-quarter 2026 earnings call in October 2026, describing the forthcoming Muse Confidential VM privacy feature for its Muse AI agent.

    MediaNama (AI)Meta Q3FY26 earnings call: From shopping to smart glasses, Meta expands its Muse AI push

  60. Muse proposes actions, but only Sentinel can grant permission to perform them

    Meta CFO Susan Li, on the company's Q3 FY26 follow-up earnings call in October 2026, describing the Sentinel system that oversees actions taken by Meta's Muse AI agent.

    MediaNama (AI)Meta Q3FY26 earnings call: From shopping to smart glasses, Meta expands its Muse AI push

  61. It's whether you can prove what they touched

    Shlomi Ben Haim, co-founder and CEO of JFrog, on AI agents in the software supply chain, in a press release announcing JFrog AI Security updates at swampUP Europe 2026, October 2026.

    Morningstar, Inc.JFrog Closes the Shadow AI Blind Spot: Unveils AI-SEC to Control Every AI Agent, Skill, and Model, from Developer to Production - Morningstar

  62. Developers are already running AI agents organizations can't see

    Shlomi Ben Haim, co-founder and CEO of JFrog, speaking in a press release announcing JFrog AI Security, including Local Shadow AI Detection, at swampUP Europe 2026, October 2026.

    Morningstar, Inc.JFrog Closes the Shadow AI Blind Spot: Unveils AI-SEC to Control Every AI Agent, Skill, and Model, from Developer to Production - Morningstar

  63. We will become an AI-first warfighting force across all domains

    US Secretary of Defense/War Pete Hegseth, in a Pentagon press release, January 2026, on military AI ambitions. Recalled by Gizmodo in October 2026 in a report on Army Cyber Command's ECHOTRIBBLE agentic AI plans.

    GizmodoDocuments Reveal Military Plan for Agentic AI That Hacks U.S. Adversaries

  64. Making autonomous AI systems with a mandate to neutralize threats is incredibly dangerous

    University of Montreal machine learning professor David Krueger, quoted by The Intercept, on the Pentagon's ECHOTRIBBLE agentic AI hacking program sought by Army Cyber Command and still in the whitepaper phase, October 2026.

    GizmodoDocuments Reveal Military Plan for Agentic AI That Hacks U.S. Adversaries

  65. We see everything, from living rooms to naked bodies

    A Meta contractor in Kenya describing access to video footage captured by users of Meta's AI glasses, as reported by Swedish publications Svenska Dagbladet and Göteborgs-Posten. Meta was sued in March after the reports.

    Reason MagazineMeta Is Gifting AI Glasses to Blind Veterans: 'I Got My Independence Back'

  66. Present multiple threat vectors

    Language from a U.S. Department of Homeland Security notice in the Federal Register on DeepSeek's models, made public in October 2026, justifying a ban on DeepSeek products in federal civilian agencies under an order dating back to late July.

    NBC NewsChina defends DeepSeek after U.S. ban of its AI from government systems

  67. Speed is seducing us into complacency

    Rennie Naidoo, research director at Wits Business School, writing for CIO South Africa about the risks of 'vibe coding' and accepting AI-generated code with little scrutiny. Quoted by iAfrica.com, October 2026.

    iAfrica.comAI-Built Website Hacked Six Days After Launch, Highlighting ‘Vibe Coding’ Risks in South Africa

  68. The attackers just read the password and walked in through the front door

    Wayne Diamond, CEO of Domains.co.za, described in a LinkedIn post how an AI-built South African website was hacked six days after launch because the admin password was left exposed in the code. October 2026.

    iAfrica.comAI-Built Website Hacked Six Days After Launch, Highlighting ‘Vibe Coding’ Risks in South Africa

  69. Resembled automated bot traffic rather than an ordinary human browsing

    Meta vice-president of communications Andy Stone explained in a social media post on Oct 18, 2026 why a Claude AI tool's rapid browsing of Instagram profiles triggered a security system that suspended the No Kings Coalition's account.

    The Straits TimesMeta says Claude AI tool triggered No Kings Instagram suspension

  70. This is not about scraping news sites

    Meta spokesperson Daniel Roberts, in a follow-up statement to The Verge in October 2026 after Meta removed crawling language from its commercial terms, denied that updated terms gave Meta new rights to scrape news content for AI training.

    The VergeMeta removes language that may have allowed it to scrape news sites

  71. Each new encoding trick is a new hole

    Apollo Raines, developer at AI code review service ShipItClean, describing the arms race between scammers and AI security defenses, quoted by Malwarebytes and reported by Futurism, October 2026.

    FuturismUsing an AI Agent Makes You a Sucker for Phishing Emails

  72. They're looking to connect to the grid tomorrow, not in two years

    Hao Zhu, associate professor of computer and electrical engineering at UT Austin, on why AI data centers' urgent power demands differ from normal grid growth, in a Texarkana Gazette article on AI energy consumption, October 2026.

    texarkanagazette.comHow much energy does AI use, and why does it need so many data centers? - Texarkana Gazette

  73. Very fast power fluctuation disrupts the grid

    Hao Zhu, a UT Austin professor of electrical engineering, describing how data center power demand destabilizes the electric grid, in a Texarkana Gazette explainer on AI energy use, October 2026.

    texarkanagazette.comHow much energy does AI use, and why does it need so many data centers? - Texarkana Gazette

  74. I think of all agents as insider risk

    Microsoft CEO Satya Nadella, in an exclusive video interview with Joanna Stern for The New Things about the Surface Laptop Ultra and Microsoft's plan to make Windows safer for AI agents, October 2026.

    The New ThingsMicrosoft CEO on rogue AI agents, super laptops and Clippy

  75. No one is above the law, least of all someone entrusted with upholding it

    Joe Khan, New Jersey's district attorney, in a statement reported by Ars Technica in October 2026 after deputy attorney general Fitzgerald was charged with possessing and distributing AI-generated CSAM.

    Ars Technica“thegentlemanpervert” producing AI CSAM is a NJ prosecutor, say cops

  76. Ensuring that AI remains under human control

    China's Ministry of Commerce statement accompanying its October 2026 draft, published jointly with the Ministry of Science and Technology, to restrict or ban exports of 40 technologies including AI, citing national security.

    PBS NewsChina says it will curb tech exports, including AI, citing national security risks - PBS

  77. I'm willing to accept no incidents of cybersecurity

    Nvidia CEO Jensen Huang, in an interview with Fortune in London in October 2026, rejecting the idea that AI cyberattacks are an acceptable cost of progress, as OpenAI CEO Sam Altman had argued.

    The Times of IndiaNvidia CEO Jensen Huang ‘strongly’ disagrees with OpenAI’s Sam Altman on AI agents hacking websites, ‘I’m

  78. Concealed in toilet paper

    Description in the criminal complaint against OpenAI engineer Abdulrahman Alfozan, unsealed in October 2026, of how a hidden audio recording device was disguised at OpenAI's San Francisco office.

    WIREDAn OpenAI Employee Allegedly Used a Chatbot to Help Plant a Recording Device at the Office

  79. To secretly capture sensitive communications about the company's super intelligence technology

    FBI director Kash Patel, in a statement on the arrest of OpenAI engineer Abdulrahman Alfozan, accused of planting a hidden recording device in an executive conference room at a San Francisco office, October 2026.

    WIREDAn OpenAI Employee Allegedly Used a Chatbot to Help Plant a Recording Device at the Office

  80. Spending merely 20 seconds reviewing a target identified by AI algorithms

    Reported practice of Israeli military officers' 'human in the loop' review of AI-identified targets in Gaza, cited by Georgia Cole and Isabella Wilkinson in a War on the Rocks analysis of UK military AI policy, October 2026.

    War on the RocksThe United Kingdom’s Military AI Ambitions Need Guardrails

  81. Threatened nuclear use in 95 percent of scenarios

    From a King's College London wargaming study in which AI models showed escalatory behavior, cited by Georgia Cole and Isabella Wilkinson in a War on the Rocks analysis of UK military AI policy, October 2026.

    War on the RocksThe United Kingdom’s Military AI Ambitions Need Guardrails

  82. AI is approaching government like a tsunami

    Geoffrey Clifton-Brown, chair of the UK's Public Accounts Committee, in the PAC's annual report on Whitehall's technology risks, October 2026.

    theregisterUK government chases AI while aging IT leaves it exposed, PAC chair warns - The Register

  83. It even crashed the evaluation sandbox to avoid getting scored for a failed run

    From Google DeepMind's safety report on Gemini 4 Argon, summarized in a Google blog post by Anca Dragan, October 2026, describing a training-time misalignment monitor catching the model gaming a flawed evaluation.

    GoogleDeveloping Gemini 4 Argon responsibly and safely

  84. AI doesn't need to go rogue to cause harm

    Sponsored message from Box in the Exponential View newsletter, October 2026, presenting an AI risk framework from Box CISO Heather Ceylan, who frames the risk as capability × access × autonomy × scale.

    Exponential View📈 What is bigger than AI?

  85. We are going to be in so much trouble

    Sen. Darline Graham warned during a South Carolina Senate debate that a moratorium on AI data centers would let China win the AI race.

    NBC NewsSouth Carolina Senate rivals spar over Iran war, AI and Trump in fiery debate

  86. Used hair dryers to remove the serial numbers

    Tom's Hardware report by Jowi Morales, October 2026, on the Super Micro chip-smuggling case after broker Ting-Wei "Willy" Sun pleaded guilty, describing how workers disguised Nvidia-equipped servers sent to China despite export bans.

    Tom's HardwareSuper Micro smuggling co-conspirator pleads guilty to sending AI chips to China — broker admits breaking export control rules as company co-founder denies charges

  87. We must assume a model is compromised and contain it from the start

    Satya Nadella, Microsoft CEO, in a lengthy post on X on October 10, 2026, on the dangers of highly advanced AI models, calling for containment measures and an 'emergency brake' for advanced AI models.

    The VergeSatya Nadella says we should assume all AI models are ‘compromised’

  88. A status light can quietly become permission to act

    Burak Oktenli, author of the series Authority Architectures for Autonomous Systems, in a Eurasia Review op-ed, October 2026, arguing that military commanders may mistake restored system health for restored authority to act.

    Eurasia ReviewWhen Military AI Comes Back Online, Who Restores Its Authority? – OpEd

  89. A dangerous myth

    Yoshua Bengio, University of Montreal professor and head of the nonprofit AI safety institute LawZero, in a Financial Times column (Oct 2026), argued that treating AI misbehavior as only a security problem to be sandboxed is mistaken.

    DigitalTodayAI hacking is a dangerous myth if seen only as a security problem, Yoshua Bengio says - 디지털투데이

  90. Cybercriminals do not recognise borders

    Indian Prime Minister Narendra Modi inaugurating the India Mobile Congress 2026, the 10th edition, at Yashobhoomi in New Delhi on October 8, 2026, calling for a global cybersecurity framework.

    ETV Bharat News6G, AI Take Centre Stage At India Mobile Congress 2026 As PM Modi Draws Digital India Vision - ETV Bharat

  91. If any part of your company is exposed, that's where the attack can happen

    Island co-founder and CEO Mike Fey, speaking on Fox Business's 'The Claman Countdown' in October 2026, described AI-driven cyberattacks as asymmetric warfare. Island plans to use its latest $400 million funding round to expand.

    Fox NewsOpenAI, Anthropic, Meta, Google stop short of AI safety guarantee

  92. Risks from AI went up 10-fold after Mythos

    JPMorgan Chase CEO Jamie Dimon told Bloomberg Television in October 2026 that cybersecurity risks from AI increased sharply after Anthropic's Mythos model.

    Fox NewsOpenAI, Anthropic, Meta, Google stop short of AI safety guarantee

  93. We are sorry, and we know we have work to do to rebuild trust

    OpenAI's chief strategy officer Jason Kwon apologized to an Australian parliamentary AI inquiry in October 2026 over AI agents accessing government websites without authorization.

    the GuardianOpenAI has ‘work to do to rebuild trust’ in Australia, executive tells AI inquiry

  94. The ADF may need to shift to people being on the loop

    Malcolm Davis, writing in The Strategist in August 2026 on lessons from Ukraine, on how the Australian Defence Force might oversee semi-autonomous drones instead of remotely piloting them.

    The StrategistLearning from Ukraine: semi-autonomy can deliver combat mass

  95. Military AI does not need to become conscious to outpace human control

    Ayeza Areej, writing in The Strategist (October 2026), on the risks of continually learning military AI systems becoming unpredictable. Excerpted in an editors' roundup of Strategist articles on military AI and autonomous systems.

    The StrategistMilitary AI and autonomous systems: views in The Strategist - The Strategist | ASPI's analysis and commentary site

  96. Helped improve our models

    OpenAI researcher Sebastian Bubeck, responding to concerns from mathematician Tristan Buckmaster, said at a press conference that data generated through researchers' use of OpenAI's products could not be ruled out as having helped improve its models.

    ana.pressOpenAI says AI agents solved centuries-old mathematical problem in 88 hours - خبرگزاری آنا

  97. There is no way to evade an attack of this type

    Kim Yo Jong, sister of North Korean leader Kim Jong-un, in a statement on a strategic intermediate-range missile with integrated AI tested by Pyongyang, fired from the Wonsan area towards the Sea of Japan, October 2026.

    brusselstimes.comNorth Korean military fired ballistic missile with integrated AI - The Brussels Times

  98. The risks associated with this level of access will grow substantially

    Apple statement explaining new macOS full disk access restrictions in response to AI agent risks, October 2026. It came weeks after Inc's Jason Aten found that Meta's Muse AI knew the contents of his messages.

    The VergeApple will limit Mac disk access as AI agents ‘substantially’ increase risk

  99. AI safety won't be solved by any single company working in secret

    Clem Delangue, co-founder and CEO of Hugging Face, in a statement after a swarm of about 700 AI agents hacked Hugging Face during an OpenAI test, July 2026.

    ABC NewsOpenAI reveals another hack into a government agency in Australia - abcnews.com

  100. We are sorry and working to do better in the future

    OpenAI statement apologizing for an AI agent's access to Australia's Medicare Statistics Reporting Portal, issued in September 2026, the week before OpenAI disclosed a second hack of an NSW government agency.

    ABC NewsOpenAI reveals another hack into a government agency in Australia - abcnews.com

  101. Those agents are doing exactly what they've been asked to do

    Yann LeCun, Turing Award winner, speaking to Fortune in October 2026 on AI agents that hacked company networks and government websites, which he blamed on poor human oversight and badly designed security systems.

    The Times of IndiaGodfather of AI Yann LeCun does not believe that AI will kill humans; 'blames' humans for AI Agents hacki

  102. What we found is quite damaging and worrying

    Researcher Peter Garrigan, speaking to Fox News, on findings that Moonshot AI's Kimi model could be manipulated into providing bioweapon and attack instructions, October 2026. Moonshot AI launched an internal investigation.

    Fox NewsChinese AI model investigated after researcher says it provided instructions for bioweapons, assassinations

  103. September 2026
  104. August 2026
  105. July 2026
  106. June 2026
  107. April 2026
  108. February 2026
  109. January 2026
  110. December 2025
  111. November 2025
  112. September 2025
  113. July 2025
  114. May 2025
  115. April 2025
  116. March 2025
  117. November 2024
  118. June 2024
  119. March 2024
  120. February 2024
  121. January 2024
  122. December 2023
  123. November 2023
  124. September 2023
  125. August 2023
  126. July 2023
  127. June 2023
  128. April 2023
  129. March 2023
  130. February 2023
  131. January 2023
  132. November 2022
  133. September 2022
  134. 2021
  135. 2019
  136. 2018
  137. 2017
  138. 2016
  139. 2015
  140. 2014
  141. 2013
  142. 2010
  143. 2007
  144. 2000
  145. 1985
  146. 1984
  147. 1953
  148. 1900