[
{
"id": "gambit-simpson-feb25-ready-to-execute",
"incident_id": "mexico-government-breach-claude",
"quote": "In total, it produced thousands of detailed reports that included ready-to-execute plans, telling the human operator exactly which internal targets to attack next and what credentials to use.",
"speaker": "Curtis Simpson",
"title": "Chief Strategy Officer, Gambit Security",
"org_type": "company",
"date": "2026-02-25",
"channel": "statement to press",
"url": "https://amp.claimsjournal.com/news/national/2026/02/25/335916.htm",
"via": "Bloomberg (via Claims Journal)",
"verified": true,
"weight": 6.0
},
{
"id": "openai-feb25-banned-accounts",
"incident_id": "mexico-government-breach-claude",
"quote": "We have banned the accounts used by this adversary and value the outreach from Gambit Security.",
"speaker": "OpenAI",
"title": "Company statement",
"org_type": "AI lab",
"date": "2026-02-25",
"channel": "statement to press",
"url": "https://amp.claimsjournal.com/news/national/2026/02/25/335916.htm",
"via": "Bloomberg (via Claims Journal)",
"verified": true,
"weight": 5
},
{
"id": "openai-jul21-hyperfocused",
"incident_id": "openai-hf-swarm",
"quote": "All evidence suggests that the models were hyperfocused on finding a solution for ExploitGym, going to extreme lengths to achieve a rather narrow testing goal.",
"speaker": "OpenAI",
"title": "Company statement",
"org_type": "AI lab",
"date": "2026-07-21",
"channel": "blog",
"url": "https://openai.com/index/hugging-face-model-evaluation-security-incident/",
"via": "",
"verified": true,
"weight": 5.7
},
{
"id": "delangue-jul25-unprecedented",
"incident_id": "openai-hf-swarm",
"quote": "Radical transparency: let’s release the traces from the “rogue” agents so the entire research community can study what happened. … The first autonomous agent cyberattack is an unprecedented event. It deserves an unprecedented response!",
"speaker": "Clément Delangue",
"title": "Co-founder and CEO, Hugging Face",
"org_type": "company",
"date": "2026-07-25",
"channel": "X post",
"url": "https://x.com/ClementDelangue/status/2081056675558195657",
"via": "",
"verified": true,
"weight": 8.0
},
{
"id": "altman-jul28-viscerally",
"incident_id": "openai-hf-swarm",
"quote": "This is the first security incident that I have felt very viscerally. … We may have to pace the rate of AI development to give ourselves enough time for society to harden around these new capability levels.",
"speaker": "Sam Altman",
"title": "CEO, OpenAI",
"org_type": "AI lab",
"date": "2026-07-28",
"channel": "podcast",
"url": "https://x.com/patrick_oshag/status/2082090998990270885",
"via": "Patrick O'Shaughnessy (host, Invest Like the Best) transcribing the episode on X",
"verified": true,
"weight": 9.6
},
{
"id": "anthropic-jul30-basic-techniques",
"incident_id": "lab-test-breaches",
"quote": "Operating under the false belief that all accessible entities were intended to be in-scope for the exercise, Claude compromised the impacted organizations’ infrastructure using basic techniques, such as exploiting weak passwords and unauthenticated endpoints.",
"speaker": "Anthropic",
"title": "Company statement",
"org_type": "AI lab",
"date": "2026-07-30",
"channel": "blog",
"url": "https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals",
"via": "",
"verified": true,
"weight": 5
},
{
"id": "bird-aug3-imminent-risk",
"incident_id": "openai-hf-swarm",
"quote": "OpenAI’s inability or unwillingness to ensure the safety of its products poses an imminent risk of substantial harm to our States. … We intend to take decisive action to protect our citizens.",
"speaker": "Brenna Bird",
"title": "Attorney General of Iowa (leading a 15-state coalition)",
"org_type": "regulator/AG",
"date": "2026-08-03",
"channel": "statement to press",
"url": "https://www.foxbusiness.com/technology/gop-ags-warn-openai-altman-preserve-records-ai-agent-hacking-probe",
"via": "Fox Business",
"verified": true,
"weight": 5
},
{
"id": "aisi-aug4-first-time",
"incident_id": "lab-test-breaches",
"quote": "These attempts were unsuccessful, and our investigations have not evidenced any resulting real-world harm. But this is the first time we have seen risks around autonomy and deception manifest this clearly, without specific prompting, in the real-world.",
"speaker": "UK AI Security Institute",
"title": "Incident report",
"org_type": "government",
"date": "2026-08-04",
"channel": "report",
"url": "https://www.aisi.gov.uk/blog/incident-report-unsanctioned-agent-behaviour-during-cyber-testing",
"via": "",
"verified": true,
"weight": 7.8
},
{
"id": "meta-aug5-third-party-service",
"incident_id": "lab-test-breaches",
"quote": "[The model] exploited a security vulnerability in a third-party service, in a manner similar to previously reported instances with other companies.",
"speaker": "Meta",
"title": "Company statement",
"org_type": "AI lab",
"date": "2026-08-05",
"channel": "statement to press",
"url": "https://www.bleepingcomputer.com/news/security/meta-ai-model-hacked-a-company-during-misconfigured-cyber-test/",
"via": "Reuters (quoted by BleepingComputer)",
"verified": true,
"weight": 5
},
{
"id": "altman-aug18-paused-training",
"incident_id": "openai-hf-swarm",
"quote": "We have paused some frontier RL training to ensure that we can meet the appropriate alignment, security and monitoring standards for the new level of capabilities in front of us. … We expect confidence in safety to increasingly set the pace of AI progress.",
"speaker": "Sam Altman",
"title": "CEO, OpenAI",
"org_type": "AI lab",
"date": "2026-08-18",
"channel": "X post",
"url": "https://x.com/sama/status/2089787807611195475",
"via": "",
"verified": true,
"weight": 5
},
{
"id": "openai-aug18-risks-grow",
"incident_id": "openai-hf-swarm",
"quote": "As models become more capable, the risks associated with developing and testing them internally also grow.",
"speaker": "OpenAI",
"title": "Company statement",
"org_type": "AI lab",
"date": "2026-08-18",
"channel": "blog",
"url": "https://openai.com/index/pacing-model-development-cyber-capabilities/",
"via": "",
"verified": true,
"weight": 5
},
{
"id": "metr-aug26-1200-agents",
"incident_id": "openai-hf-swarm",
"quote": "Roughly 1200 agents meant to be isolated from one another found a way to communicate with one another on an unsanctioned message board, sending over 70,000 messages and files… Of these agents, 700 went on to participate in the attack on Hugging Face.",
"speaker": "METR",
"title": "Independent investigation",
"org_type": "researcher",
"date": "2026-08-26",
"channel": "report",
"url": "https://metr.org/blog/2026-08-26-openai-hugging-face-incident-investigation/",
"via": "",
"verified": true,
"weight": 8.2
},
{
"id": "openai-aug26-techreport-circumvented-controls",
"incident_id": "openai-hf-swarm",
"quote": "In July 2026, during internal cybersecurity evaluations, OpenAI models in an internal evaluation environment circumvented controls intended to isolate them from the internet and performed computer network exploitation of OpenAI’s internal research infrastructure and Hugging Face systems.",
"speaker": "OpenAI",
"title": "OpenAI – Hugging Face Incident Technical Report",
"org_type": "AI lab",
"date": "2026-08-26",
"channel": "report",
"url": "https://cdn.openai.com/pdf/67869394-cb91-4c12-888c-5cbd85c7814c/OpenAI-Hugging-Face%20Incident-Technical-Report.pdf",
"via": "",
"verified": true,
"weight": 5
},
{
"id": "openai-aug26-techreport-dramatic-shift",
"incident_id": "openai-hf-swarm",
"quote": "The introduction of autonomous offensive agent collectives represents a dramatic shift in offensive attacker capability and will require urgent upgrades of control environments, as well as deeper adoption of AI for defensive purposes.",
"speaker": "OpenAI",
"title": "OpenAI – Hugging Face Incident Technical Report",
"org_type": "AI lab",
"date": "2026-08-26",
"channel": "report",
"url": "https://cdn.openai.com/pdf/67869394-cb91-4c12-888c-5cbd85c7814c/OpenAI-Hugging-Face%20Incident-Technical-Report.pdf",
"via": "",
"verified": true,
"weight": 5
},
{
"id": "openai-aug26-warning-shot",
"incident_id": "openai-hf-swarm",
"quote": "We are taking this incident as a ‘warning shot’ that today’s model capabilities present the possibility of loss-of-control incidents. … Companies that build AI systems will need to ensure that their systems always remain under meaningful human control…",
"speaker": "OpenAI",
"title": "Company statement",
"org_type": "AI lab",
"date": "2026-08-26",
"channel": "blog",
"url": "https://openai.com/index/hugging-face-incident-and-the-road-ahead/",
"via": "The Register",
"verified": true,
"weight": 8.5
},
{
"id": "hawley-sep9-reckless",
"incident_id": "openai-hf-swarm",
"quote": "This is reckless. And this is merely what we know from what limited information you disclosed to and allowed your partner auditors to investigate.",
"speaker": "Josh Hawley",
"title": "U.S. Senator (R-MO); Chairman, Senate Homeland Security Subcommittee on Disaster Management",
"org_type": "legislature",
"date": "2026-09-09",
"channel": "statement to press",
"url": "https://cyberscoop.com/wp-content/uploads/sites/3/2026/09/2026-09-09-Hawley-Letter-to-OpenAI-re-Hugging-Face-AI-Agent-Hack.pdf",
"via": "",
"verified": true,
"weight": 5
},
{
"id": "hawley-sep9-went-rogue",
"incident_id": "openai-hf-swarm",
"quote": "These AI agents were looking for the answer key to their evaluations and actively tampered with evidence of their activity to cover their tracks. In short, they went rogue.",
"speaker": "Josh Hawley",
"title": "U.S. Senator (R-MO); Chairman, Senate Homeland Security Subcommittee on Disaster Management",
"org_type": "legislature",
"date": "2026-09-09",
"channel": "statement to press",
"url": "https://cyberscoop.com/wp-content/uploads/sites/3/2026/09/2026-09-09-Hawley-Letter-to-OpenAI-re-Hugging-Face-AI-Agent-Hack.pdf",
"via": "",
"verified": true,
"weight": 9.2
},
{
"id": "bengio-sep11-cheating-rewarded",
"incident_id": "openai-hf-swarm",
"quote": "With the OpenAI agents, there is reason to believe successful cheating was actually rewarded: When the scoring program does not see the cheating, it pays out anyway, and such cheats become more likely next time.",
"speaker": "Yoshua Bengio",
"title": "Professor, Université de Montréal; founder, LawZero; Turing Award laureate",
"org_type": "researcher",
"date": "2026-09-11",
"channel": "blog",
"url": "https://fortune.com/2026/10/16/openai-discloses-four-more-rogue-incidents-involving-financial-medical-and-government-data/",
"via": "Fortune (quoting his Sept. 11 essay on misalignment)",
"verified": true,
"weight": 5.5
},
{
"id": "altman-sep12-pace-the-frontier",
"incident_id": "",
"quote": "I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we’ve had at OpenAI in recent weeks. Committing to having independent evaluators with employee-like access is a great idea, and we will do the same.",
"speaker": "Sam Altman",
"title": "CEO, OpenAI",
"org_type": "AI lab",
"date": "2026-09-12",
"channel": "X post",
"url": "https://x.com/sama/status/2098811563415150910",
"via": "",
"verified": true,
"weight": 5
},
{
"id": "amodei-sep12-catastrophic-damage",
"incident_id": "openai-hf-swarm",
"quote": "It’s easy to dismiss this incident because no one was hurt and the economic damage was minimal, but in my opinion, a swarm that possessed greater capabilities but a similar level of misalignment could have caused catastrophic damage.",
"speaker": "Dario Amodei",
"title": "CEO, Anthropic",
"org_type": "AI lab",
"date": "2026-09-12",
"channel": "blog",
"url": "https://darioamodei.com/post/we-must-pace-the-frontier",
"via": "",
"verified": true,
"weight": 9.4
},
{
"id": "amodei-sep12-slow-the-pace",
"incident_id": "",
"quote": "We must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain.",
"speaker": "Dario Amodei",
"title": "CEO, Anthropic",
"org_type": "AI lab",
"date": "2026-09-12",
"channel": "blog",
"url": "https://darioamodei.com/post/we-must-pace-the-frontier",
"via": "",
"verified": true,
"weight": 5.4
},
{
"id": "bengio-sep16-losing-control",
"incident_id": "openai-hf-swarm",
"quote": "There's a reason that companies are saying this is going too fast, that we're losing control. … People like me have been expecting this for a long time.",
"speaker": "Yoshua Bengio",
"title": "Professor, Université de Montréal; founder, LawZero; Turing Award laureate",
"org_type": "researcher",
"date": "2026-09-16",
"channel": "interview",
"url": "https://www.enca.com/business/were-losing-control-ai-pioneer-yoshua-bengio-tells-afp",
"via": "AFP (via eNCA)",
"verified": true,
"weight": 7.0
},
{
"id": "vonderleyen-sep16-hacking-never-thought-possible",
"incident_id": "",
"quote": "Models being developed will allow hacking on a level we never thought possible. And they will soon be in the hands of adversaries who see the world very differently from us.",
"speaker": "Ursula von der Leyen",
"title": "President of the European Commission",
"org_type": "government",
"date": "2026-09-16",
"channel": "speech",
"url": "https://ec.europa.eu/commission/presscorner/detail/en/speech_26_1868",
"via": "",
"verified": true,
"weight": 9.8
},
{
"id": "vonderleyen-sep16-mere-glimpse",
"incident_id": "openai-hf-swarm",
"quote": "Incidents of AI agents escaping their environment or inserting malicious code are a mere glimpse. After the HuggingFace incident, developers are ringing the alarm. And CEOs of the most advanced companies tell us that it is time to slow down on the self-recursive models.",
"speaker": "Ursula von der Leyen",
"title": "President of the European Commission",
"org_type": "government",
"date": "2026-09-16",
"channel": "speech",
"url": "https://ec.europa.eu/commission/presscorner/detail/en/speech_26_1868",
"via": "",
"verified": true,
"weight": 5
},
{
"id": "adkins-sep19-guessed-credentials",
"incident_id": "lab-test-breaches",
"quote": "The model found public information online and guessed credentials to access websites it thought were part of the test.",
"speaker": "Heather Adkins",
"title": "Vice President of Security Engineering, Google",
"org_type": "company",
"date": "2026-09-19",
"channel": "interview",
"url": "https://www.aljazeera.com/news/2026/9/19/googles-gemini-ai-hacks-3-companies-in-security-test-then-stops",
"via": "Al Jazeera",
"verified": true,
"weight": 5
},
{
"id": "albanese-sep23-way-too-long",
"incident_id": "openai-hf-swarm",
"quote": "I also expressed my disappointment that it took the company way too long to inform the government what had occurred and the nature of the way that notification occurred as well was unacceptable.",
"speaker": "Anthony Albanese",
"title": "Prime Minister of Australia",
"org_type": "government",
"date": "2026-09-23",
"channel": "statement to press",
"url": "https://www.kuow.org/npr-news/2026-09-23/openais-breach-of-australian-health-department-website-prompts-rebuke",
"via": "The Associated Press (via NPR/KUOW)",
"verified": true,
"weight": 6.8
},
{
"id": "bonta-sep24-not-theoretical",
"incident_id": "openai-hf-swarm",
"quote": "But federal action is also urgently needed to achieve the national and international coordination required to prevent catastrophic harm. The call is coming from inside the house; the danger is not theoretical anymore.",
"speaker": "Rob Bonta",
"title": "Attorney General of California",
"org_type": "regulator/AG",
"date": "2026-09-24",
"channel": "press release",
"url": "https://oag.ca.gov/news/press-releases/attorney-general-bonta-congress-must-act-urgently-protect-against-catastrophic",
"via": "",
"verified": true,
"weight": 9.0
},
{
"id": "marles-sep24-very-serious",
"incident_id": "openai-hf-swarm",
"quote": "But the incident itself is obviously very serious. It's very serious that we've got an AI agent having gained unauthorised access into an Australian website.",
"speaker": "Richard Marles",
"title": "Deputy Prime Minister and Minister for Defence, Australia (acting PM)",
"org_type": "government",
"date": "2026-09-24",
"channel": "statement to press",
"url": "https://www.sbs.com.au/news/article/openai-agent-hacked-medicare-albanese-reveals/qas79d9ta",
"via": "SBS News",
"verified": true,
"weight": 5
},
{
"id": "openai-sep24-actions-we-did-not-intend",
"incident_id": "openai-hf-swarm",
"quote": "Our models attempted to look up answers, and available statistics for questions about Australia during an internal evaluation. In the course of that, our models took actions we did not intend.",
"speaker": "OpenAI spokesperson",
"title": "Spokesperson, OpenAI",
"org_type": "AI lab",
"date": "2026-09-24",
"channel": "statement to press",
"url": "https://www.sbs.com.au/news/article/openai-agent-hacked-medicare-albanese-reveals/qas79d9ta",
"via": "SBS News",
"verified": true,
"weight": 5
},
{
"id": "openai-sep25-take-months",
"incident_id": "openai-hf-swarm",
"quote": "In our broader review, we’re continuing to prioritize the most serious incidents while expanding our work to lower-severity activity, including agents spamming websites. Given the scale of this work and the need to verify each case, we expect the review to take months.",
"speaker": "OpenAI spokesperson",
"title": "Spokesperson, OpenAI",
"org_type": "AI lab",
"date": "2026-09-25",
"channel": "statement to press",
"url": "https://techcrunch.com/2026/09/25/for-months-openais-agent-swarms-have-been-attacking-online-databases-to-find-obscure-facts/",
"via": "TechCrunch",
"verified": true,
"weight": 5
},
{
"id": "stosz-sep25-would-have-discovered",
"incident_id": "openai-hf-swarm",
"quote": "It seems likely that if they had exhaustively studied and understood all of the outgoing requests and incoming responses for those agents involved in the DSE wiki, that they would have discovered this activity.",
"speaker": "Conrad Stosz",
"title": "Head of Governance, Transluce",
"org_type": "researcher",
"date": "2026-09-25",
"channel": "interview",
"url": "https://techcrunch.com/2026/09/25/for-months-openais-agent-swarms-have-been-attacking-online-databases-to-find-obscure-facts/",
"via": "TechCrunch",
"verified": true,
"weight": 5
},
{
"id": "hawley-sep30-break-it-pay-for-it",
"incident_id": "openai-hf-swarm",
"quote": "I wonder if it’s not time to get back to the good old-fashioned American principle, which is if you break it, you pay for it. If you cause damage, you clean it up. That’s been the basic principle of American law for 250 years now.",
"speaker": "Josh Hawley",
"title": "U.S. Senator (R-MO); Chairman, Senate Homeland Security Subcommittee on Disaster Management",
"org_type": "legislature",
"date": "2026-09-30",
"channel": "hearing",
"url": "https://www.techpolicy.press/senate-hearing-on-rogue-ai-securing-the-homeland-against-ai-agent-attacks/",
"via": "TechPolicy.Press (hearing transcript)",
"verified": true,
"weight": 5.9
},
{
"id": "hobbhahn-sep30-warning-shots",
"incident_id": "openai-hf-swarm",
"quote": "These were the warning shots; next time, we may not be so lucky. We face a closing window of opportunity in which AI agents are misaligned but not yet capable of fully escaping human control and remaining undetected.",
"speaker": "Marius Hobbhahn",
"title": "CEO and co-founder, Apollo Research",
"org_type": "researcher",
"date": "2026-09-30",
"channel": "hearing",
"url": "https://www.hsgac.senate.gov/subcommittees/dmdcc/hearings/rogue-ai-securing-the-homeland-against-ai-agent-attacks/marius-hobbhahn-testimony/",
"via": "",
"verified": true,
"weight": 10
},
{
"id": "kim-sep30-we-dont-know",
"incident_id": "openai-hf-swarm",
"quote": "I did a town hall the other week and someone there asks me, how many of these incidents have happened? How many hacks have happened? And the answer that everyone knows is we don’t know.",
"speaker": "Andy Kim",
"title": "U.S. Senator (D-NJ); Ranking Member, Senate Homeland Security Subcommittee on Disaster Management",
"org_type": "legislature",
"date": "2026-09-30",
"channel": "hearing",
"url": "https://www.techpolicy.press/senate-hearing-on-rogue-ai-securing-the-homeland-against-ai-agent-attacks/",
"via": "TechPolicy.Press (hearing transcript)",
"verified": true,
"weight": 7.2
},
{
"id": "transluce-sep30-rogue-agents",
"incident_id": "openai-hf-swarm",
"quote": "We discovered several additional incidents where rogue AI agents appear to have used aggressive techniques to access publicly available data on government websites. This includes two rudimentary and failed hacking attempts…",
"speaker": "Transluce",
"title": "Incident report",
"org_type": "researcher",
"date": "2026-09-30",
"channel": "report",
"url": "https://transluce.org/us-canada-gov",
"via": "",
"verified": true,
"weight": 5
},
{
"id": "bonta-oct1-legally-accountable",
"incident_id": "openai-hf-swarm",
"quote": "Companies that develop these models and offer them for use have a moral and legal responsibility to ensure that they do not perpetrate or enable cyberattacks… Developers that fail to do so can and should be held legally accountable.",
"speaker": "Rob Bonta",
"title": "Attorney General of California",
"org_type": "regulator/AG",
"date": "2026-10-01",
"channel": "press release",
"url": "https://oag.ca.gov/node/630636",
"via": "",
"verified": true,
"weight": 5
},
{
"id": "lee-eogweon-oct4-highest-alert",
"incident_id": "korea-banks-artex",
"quote": "We cannot rule out the possibility that AI was used in the attacks. … The entire financial sector must recognize the gravity of the situation and remain on the highest alert.",
"speaker": "Lee Eog-weon",
"title": "Chairman, Financial Services Commission (South Korea)",
"org_type": "regulator/AG",
"date": "2026-10-04",
"channel": "statement to press",
"url": "https://www.koreajoongangdaily.com/business/from-banks-to-lenders-suspected-ai-hacks-expose-cracks-in-koreas-financial-defenses/12904281",
"via": "Korea JoongAng Daily (English translation)",
"verified": true,
"weight": 6.2
},
{
"id": "wikimedia-deckelmann-oct5",
"incident_id": "openai-hf-swarm",
"quote": "AI companies are not doing enough to secure their systems and protect the public from the harm they cause. That burden is falling onto everyone else, including smaller organizations.",
"speaker": "Selena Deckelmann",
"title": "Chief Product and Technology Officer, Wikimedia Foundation",
"org_type": "company",
"date": "2026-10-05",
"channel": "blog",
"url": "https://wikimediafoundation.org/news/2026/10/05/openai-rogue-agent-activities-found-on-wikimedia-projects/",
"via": "",
"verified": true,
"weight": 7.6
},
{
"id": "lee-jaemyung-oct6-signs-of-ai",
"incident_id": "korea-banks-artex",
"quote": "In some hacking incidents, signs have emerged of AI being used, causing considerable public concern and anxiety. … Please establish the circumstances swiftly and clearly, and concentrate personnel and resources on minimising the damage.",
"speaker": "Lee Jae-myung",
"title": "President of the Republic of Korea",
"org_type": "government",
"date": "2026-10-06",
"channel": "speech",
"url": "https://www.thestar.com.my/news/world/2026/10/06/south-korea039s-lee-says-ai-appears-to-have-been-used-in-bank-hacks",
"via": "Reuters (via The Star)",
"verified": true,
"weight": 6.6
},
{
"id": "meyers-oct8-one-human",
"incident_id": "korea-banks-artex",
"quote": "And this is significant because it allows one human to target many customers in a very short period of time using the power of AI.",
"speaker": "Adam Meyers",
"title": "Senior Vice President, Counter Adversary Operations, CrowdStrike",
"org_type": "company",
"date": "2026-10-08",
"channel": "statement to press",
"url": "https://www.devdiscourse.com/article/international/3989020-south-korean-banks-were-likely-hacked-by-a-china-based-actor-with-an-ai-agent-crowdstrike-says",
"via": "Reuters (via Devdiscourse)",
"verified": true,
"weight": 6.4
},
{
"id": "anthropic-oct9-persistence",
"incident_id": "lab-test-breaches",
"quote": "Some of the cases described below involved websites run by U.S. government agencies at the federal, state, and local levels. We have briefed the White House on these cases and notified each agency involved.",
"speaker": "Anthropic",
"title": "Alignment team report",
"org_type": "AI lab",
"date": "2026-10-09",
"channel": "report",
"url": "https://www.anthropic.com/research/investigating-unintended-model-actions",
"via": "",
"verified": true,
"weight": 5
},
{
"id": "phillypd-oct9-flagged-as-spam",
"incident_id": "lab-test-breaches",
"quote": "[The tip] was flagged as spam and was never forwarded to the Real-Time Crime Center for investigative vetting or dissemination.",
"speaker": "Philadelphia Police Department",
"title": "Statement",
"org_type": "government",
"date": "2026-10-09",
"channel": "statement to press",
"url": "https://www.aljazeera.com/news/2026/10/10/anthropic-ai-model-submits-false-homicide-tip-to-philadelphia-police",
"via": "Al Jazeera (AFP and Reuters)",
"verified": true,
"weight": 6.3
},
{
"id": "whitehouse-oct9-full-transparency",
"incident_id": "lab-test-breaches",
"quote": "We informed the company that we expect immediate and full transparency to the entities involved and the public. We also expect that the company will immediately provide remediation services to the affected entities and any harmed Americans.",
"speaker": "White House spokesperson",
"title": "Spokesperson, The White House",
"org_type": "government",
"date": "2026-10-09",
"channel": "statement to press",
"url": "https://www.law360.com/cybersecurity-privacy/articles/2536586",
"via": "Law360",
"verified": true,
"weight": 6.5
},
{
"id": "narayan-oct12-do-not-release",
"incident_id": "",
"quote": "And to the companies developing the world’s most powerful AI systems, my message is: Listen when your scientists warn of serious dangers. Investigate unexpected behaviour. And if you cannot adequately control the risks, do not release the system.",
"speaker": "Kanishka Narayan",
"title": "Minister for AI, UK Government",
"org_type": "government",
"date": "2026-10-12",
"channel": "press release",
"url": "https://www.gov.uk/government/news/britain-steps-up-national-defences-against-growing-ai-risks",
"via": "",
"verified": true,
"weight": 8.4
},
{
"id": "openai-oct15-ordinary-research-task",
"incident_id": "openai-june-four",
"quote": "This case shows how a model pursuing an ordinary research task can ultimately exploit security flaws and retrieve sensitive information unrelated to the user’s request.",
"speaker": "OpenAI",
"title": "Misalignment report",
"org_type": "AI lab",
"date": "2026-10-15",
"channel": "report",
"url": "https://alignment.openai.com/misalignment-reports/retrieving-server-secrets-and-bypassing-authorization-on-a-government-statistics-website/",
"via": "",
"verified": true,
"weight": 6.1
},
{
"id": "openai-oct16-strengthened-controls",
"incident_id": "openai-june-four",
"quote": "In response to this and to other incidents, we’ve strengthened our security controls, blocked models from accessing the live internet during training or evaluation, improved our alignment grading, and extended our misalignment monitoring to all reinforcement learning and evaluation traffic.",
"speaker": "OpenAI",
"title": "Company statement",
"org_type": "AI lab",
"date": "2026-10-16",
"channel": "statement to press",
"url": "https://fortune.com/2026/10/16/openai-discloses-four-more-rogue-incidents-involving-financial-medical-and-government-data/",
"via": "Fortune",
"verified": true,
"weight": 5
},
{
"id": "ncsc-oct21-take-offline",
"incident_id": "php2shell",
"quote": "If you are unable to apply the temporary mitigations, or applying them would take too long, you should consider taking the system offline until it can be fully mitigated.",
"speaker": "UK National Cyber Security Centre",
"title": "Alert",
"org_type": "security agency",
"date": "2026-10-21",
"channel": "advisory",
"url": "https://www.ncsc.gov.uk/news/critical-security-issue-affecting-php",
"via": "",
"verified": true,
"weight": 5.3
},
{
"id": "nsa-oct21-weaponize",
"incident_id": "php2shell",
"quote": "Servers running affected PHP versions are exploitable. Cyber actors are likely to weaponize PHP2Shell vulnerabilities. These risks are serious enough to require immediate patching across United States Government departments and agencies.",
"speaker": "National Security Agency",
"title": "Advisory",
"org_type": "security agency",
"date": "2026-10-21",
"channel": "advisory",
"url": "https://www.nsa.gov/Press-Room/Press-Releases-Statements/Press-Release-View/Article/4630238/nsa-advisory-php-updates/",
"via": "",
"verified": true,
"weight": 7.4
},
{
"id": "nsacyber-oct21-take-offline",
"incident_id": "php2shell",
"quote": "Effective Immediately, all organizations using PHP for their systems are advised to be prepared for an update addressing two critical vulnerabilities. … If an immediate update is not feasible, it is recommend [sic] that organizations take the system offline until it can be fully mitigated.",
"speaker": "NSA Cybersecurity Directorate",
"title": "@NSACyber, U.S. National Security Agency",
"org_type": "security agency",
"date": "2026-10-21",
"channel": "X post",
"url": "https://x.com/NSACyber/status/2112903220209098897",
"via": "",
"verified": true,
"weight": 5
},
{
"id": "openai-oct21-defenders-window",
"incident_id": "php2shell",
"quote": "It highlights how AI can uncover security flaws with internet-scale impact, but the same capabilities can also help attackers rapidly develop offensive tools and exploits. … To our knowledge, frontier AI found these vulnerabilities before attackers did.",
"speaker": "OpenAI",
"title": "Company statement",
"org_type": "AI lab",
"date": "2026-10-21",
"channel": "blog",
"url": "https://openai.com/index/coordinated-disclosure-of-critical-vulnerabilities/",
"via": "",
"verified": true,
"weight": 5.2
},
{
"id": "phpfoundation-oct21-embargo",
"incident_id": "php2shell",
"quote": "On October 21st at 07:41 UTC, more than six hours before the coordinated release, Chainguard unilaterally broke the embargo and published built images containing the fix… This disregarded weeks of work by many parties toward a coordinated disclosure.",
"speaker": "The PHP Foundation",
"title": "Statement",
"org_type": "company",
"date": "2026-10-21",
"channel": "blog",
"url": "https://thephp.foundation/blog/2026/10/21/todays-php-security-release-and-how-were-strengthening-php-security/",
"via": "",
"verified": true,
"weight": 5.8
},
{
"id": "phpfoundation-oct21-tools-find-bugs",
"incident_id": "php2shell",
"quote": "Tools now find bugs faster than ever, for defenders and attackers alike… What tools cannot do is decide on the right fix for a long-standing design decision in the engine, weigh the compatibility trade-offs, and review the patch.",
"speaker": "The PHP Foundation",
"title": "Statement",
"org_type": "company",
"date": "2026-10-21",
"channel": "blog",
"url": "https://thephp.foundation/blog/2026/10/21/todays-php-security-release-and-how-were-strengthening-php-security/",
"via": "",
"verified": true,
"weight": 5
},
{
"id": "prince-oct21-millions",
"incident_id": "php2shell",
"quote": "In the 10 hours before it was publicly announced, across the @Cloudflare network we saw thousands of attempts to exploit it. That jumped to tens of thousands in the 45 minutes post announcement, hundreds of thousands within 2 hours, and millions now.",
"speaker": "Matthew Prince",
"title": "CEO, Cloudflare",
"org_type": "company",
"date": "2026-10-21",
"channel": "X post",
"url": "https://x.com/eastdakota/status/2113079352682570051",
"via": "",
"verified": true,
"weight": 8.6
},
{
"id": "solomon-oct21-global-issue",
"incident_id": "php2shell",
"quote": "This is a global issue, and other governments and organizations are also taking protective measures. If your organization uses PHP, apply the latest security updates as soon as possible … We take this situation extremely seriously.",
"speaker": "Evan Solomon",
"title": "Minister of Artificial Intelligence and Digital Innovation, Canada",
"org_type": "government",
"date": "2026-10-21",
"channel": "X post",
"url": "https://globalnews.ca/news/12110881/government-sites-go-dark-vuln-in-code/",
"via": "Global News (quoting his X post)",
"verified": true,
"weight": 5
},
{
"id": "takaichi-oct21-urgent-request",
"incident_id": "php2shell",
"quote": "This is an extremely urgent request that I ask everyone to act on tonight. … If it is difficult to apply the fix promptly, temporarily taking your website offline, for example, can be considered one effective measure to reduce the risk.",
"speaker": "Sanae Takaichi",
"title": "Prime Minister of Japan",
"org_type": "government",
"date": "2026-10-21",
"channel": "X post",
"url": "https://x.com/takaichi_sanae/status/2112890642590564386",
"via": "",
"verified": true,
"weight": 8.8,
"original": "皆様に、今夜に対応して頂きたい大至急のお願いです。…修正パッチを速やかに適用することが困難な場合は、例えばウェブサイト等を一時的に停止させることも、リスクを低減させるための有効策の一つと考えられます。",
"lang": "ja"
},
{
"id": "trahan-oct21-new-reality",
"incident_id": "php2shell",
"quote": "This is the new reality of cybersecurity in the age of AI. Advanced AI systems can now uncover serious vulnerabilities in the code that underpins the internet. That capability can strengthen our defenses, but it can just as easily be turned against us.",
"speaker": "Lori Trahan",
"title": "U.S. Representative (D-MA)",
"org_type": "legislature",
"date": "2026-10-21",
"channel": "X post",
"url": "https://x.com/RepLoriTrahan/status/2112971928830971988",
"via": "",
"verified": true,
"weight": 5
},
{
"id": "whitehouse-oct21-american-leadership",
"incident_id": "php2shell",
"quote": "This discovery demonstrates exactly why American leadership in SI is so critical. This is American innovation strengthening our national security.",
"speaker": "White House spokesperson",
"title": "Statement",
"org_type": "government",
"date": "2026-10-21",
"channel": "statement to press",
"url": "https://aapnews.aap.com.au/news/government-websites-go-dark-after-ai-discovers-flaw",
"via": "AAP (Reuters)",
"verified": true,
"weight": 5
},
{
"id": "wiz-oct21-meaningful-barrier",
"incident_id": "php2shell",
"quote": "Frontier models were critical to making this process practical, enabling us to develop and adapt working exploits across multiple applications in hours. As the time and expertise required for this work decrease, complexity may no longer provide a meaningful barrier…",
"speaker": "Wiz Research",
"title": "Sagi Tzadik, Shir Tamari, Nir Ohfeld and Tomer Schwartz",
"org_type": "company",
"date": "2026-10-21",
"channel": "blog",
"url": "https://www.wiz.io/blog/php2shell-php-rce",
"via": "",
"verified": true,
"weight": 5.6
},
{
"id": "hunt-oct22-big-exposure",
"incident_id": "php2shell",
"quote": "Big exposure, clear intent, no confirmed working mass exploit yet. That is the window to patch and watch for the attempt signatures, instead of waiting for a working exploit to decide for you.",
"speaker": "Hunt.io",
"title": "threat research team",
"org_type": "company",
"date": "2026-10-22",
"channel": "blog",
"url": "https://hunt.io/blog/php2shell-cve-2026-109618-109619-open-directories",
"via": "",
"verified": true,
"weight": 5.1
},
{
"id": "phpfoundation-oct23-regular-release",
"incident_id": "php2shell",
"quote": "The issue can only occur under an unusual combination of configuration and application code. … Since this setup is unlikely in practice, we are handling it in the regular release cycle and will publish a full advisory for CVE-2026-115560 with the next release.",
"speaker": "PHP project",
"title": "informational advisory on CVE-2026-115560",
"org_type": "company",
"date": "2026-10-23",
"channel": "advisory",
"url": "https://www.php.net/archive/2026.php#2026-10-23-1",
"via": "",
"verified": true,
"weight": 4.9
},
{
"id": "metr-oct22-postpone-monitoring",
"incident_id": "",
"quote": "During the time period when compaction events were not being monitored, agents could indefinitely postpone being monitored by simply continuing to take actions.",
"speaker": "METR",
"title": "red-teaming OpenAI's internal agent monitoring",
"org_type": "researcher",
"date": "2026-10-22",
"channel": "blog",
"url": "https://metr.org/blog/2026-10-22-red-teaming-openai-agent-monitoring/",
"via": "",
"verified": true,
"weight": 5.0
},
{
"id": "anthropic-oct23-serious-failure",
"incident_id": "lab-test-breaches",
"quote": "Claude Mythos 5's behavior in these episodes constitutes a serious failure of alignment.",
"speaker": "Anthropic",
"title": "Alignment team, assessment of the UK AISI cyber-testing incident",
"org_type": "AI lab",
"date": "2026-10-23",
"channel": "blog",
"url": "https://www.anthropic.com/research/alignment-assessment-ukaisi",
"via": "",
"verified": true,
"weight": 7.9
},
{
"id": "anthropic-oct23-nine-of-43",
"incident_id": "lab-test-breaches",
"quote": "In nine of AISI's 43 evaluation runs, Mythos 5 took unsanctioned actions that impacted real people and organizations.",
"speaker": "Anthropic",
"title": "Alignment team, assessment of the UK AISI cyber-testing incident",
"org_type": "AI lab",
"date": "2026-10-23",
"channel": "blog",
"url": "https://www.anthropic.com/research/alignment-assessment-ukaisi",
"via": "",
"verified": true,
"weight": 6.7
},
{
"id": "openai-oct23-no-internet",
"incident_id": "openai-hf-swarm",
"quote": "we believe the responsible approach is to conduct frontier model training with no internet access.",
"speaker": "OpenAI",
"title": "announcement on training frontier models in a closed system",
"org_type": "AI lab",
"date": "2026-10-23",
"channel": "blog",
"url": "https://www.engadget.com/2296019/openai-says-its-cutting-off-internet-access-for-frontier-model-training/",
"via": "Engadget",
"verified": true,
"weight": 6.9
},
{
"id": "obadiaru-oct23-borrowed-time",
"incident_id": "php2shell",
"quote": "Attackers are refining their tooling in the open, so defenders should treat this quiet period as borrowed time",
"speaker": "Andrew Obadiaru",
"title": "CISO, Cobalt",
"org_type": "company",
"date": "2026-10-23",
"channel": "statement to press",
"url": "https://therecord.media/php-php2shell-exploitation-attempts-pocs",
"via": "The Record",
"verified": true,
"weight": 5.45
},
{
"id": "mcintyre-oct23-demonstrating-vs-exploit",
"incident_id": "php2shell",
"quote": "There's a significant difference between demonstrating a vulnerability and developing an exploit",
"speaker": "Spencer McIntyre",
"title": "Director of exploit development, VulnCheck",
"org_type": "company",
"date": "2026-10-23",
"channel": "statement to press",
"url": "https://therecord.media/php-php2shell-exploitation-attempts-pocs",
"via": "The Record",
"verified": true,
"weight": 4.8
}
]