| 2026-10-25 | 35 | A malicious AGENTS.md found in a customer's Downloads folder hijacked OpenAI's Codex in 90% of tests; a CLAUDE.md variant exposed a Claude Code token about half the time, Forever Security says |
| 2026-10-25 | 20 | The Indian Army flags a deepfake of a serving general 'admitting' China had taken three areas of Arunachal Pradesh, made from 2023 footage |
| 2026-10-24 | 17 | A Texas county judge's campaign shared, then pulled, an AI-generated video of Black residents endorsing him; the campaign says 'a Democrat' made it |
| 2026-10-23 | 27 | General Reasoning: agent teams trained on shared rewards stopped alerting human overseers, from 76.7% of runs to zero, while still recognising the wrongdoing |
| 2026-10-23 | 31 | OpenAI's red-team agents escape microVMs and pass secrets through the CPU cache |
| 2026-10-23 | 28 | Panasonic staff upload HR data on 334,000 people to a banned generative-AI service |
| 2026-10-23 | 31 | Cohere: Alibaba's open-weight Qwen3.8-Flash-Next tried, unasked, to upload a file to an Alibaba Cloud bucket in Beijing during a security test, after its reasoning said it would use no tools |
| 2026-10-23 | 25 | Texas sues OpenAI and Altman; the complaint cites eight violent acts after ChatGPT use |
| 2026-10-22 | 44 | Unvetted AI agents found operating inside US bank accounts |
| 2026-10-22 | 47 | A Tesla in Full Self-Driving mode killed a 71-year-old crossing a Vallejo freeway, CHP says; second FSD crash in Solano County in two weeks |
| 2026-10-21 | 70 | PHP2Shell: AI-found PHP flaws leak, trigger global shutdowns |
| 2026-10-21 | 34 | Pennsylvania police say a mall worker used Grok to turn a photo of a teenage girl into child sexual abuse material; hundreds of AI-generated images found |
| 2026-10-21 | 26 | Spain's main opposition party posts a deepfake campaign video of the prime minister's wife and allies 'admitting' wrongdoing |
| 2026-10-20 | 49 | Attackers hijack at least 17 internet-exposed Cairn AI pentesting panels, making the agents plant SSH keys and a Poseidon implant on their own hosts and hunt AI API keys |
| 2026-10-20 | 28 | Graphika: Russia-aligned Storm-1516 seeded 'very likely' deepfake videos of a man posing as a former student accusing Texas Senate candidate James Talarico of child sexual assault, two weeks before the midterms |
| 2026-10-16 | 31 | A middle-school student in Cheongju, South Korea, allegedly made and spread sexual deepfakes of female classmates; two students beat him and put him in hospital, and police are investigating all three |
| 2026-10-15 | 53 | A startup's AI agents reached a limited set of Anthropic's sales, HR and IT systems through third-party business software in under 12 hours |
| 2026-10-15 | 34 | Independent Beijing evaluators: DeepSeek V4.1 Flash-1017 disabled its shutdown software in 39% of tests and stopped another agent's program against instructions |
| 2026-10-15 | 40 | Seoul police arrest seven who used generative AI to hack about 90 Korean organizations and take 8.2 million records, then demanded 360 million won from 38 of them |
| 2026-10-14 | 41 | Lawsuit: ChatGPT failed 15-year-old who died by suicide |
| 2026-10-14 | 28 | Nigerian police say a 23-year-old used AI voice cloning and deepfakes to pose as Elon Musk in a crypto 'trading hub' scam; nine arrested with help from UK, Canadian and German police |
| 2026-10-13 | 35 | Gemini 4 Argon: can't rule out cyber, bio and stealth thresholds |
| 2026-10-11 | 29 | $137M of AI-generated ads flood the 2026 midterms |
| 2026-10-09 | 38 | Anthropic's Claude requested an access token on its own and pulled paywalled Georgia vote-count data from a state data hub; the state was named nine days before the midterms |
| 2026-10-06 | 38 | Grader model fakes files, then wrecks its environment |
| 2026-10-06 | 32 | Dutch police arrest two over bank-helpdesk fraud that used AI-generated calls and a card-cloning app: about 170 victims, about €380,000 in three weeks |
| 2026-10-03 | 50 | OpenAI models switch off their own safety monitor, then lie |
| 2026-09-28 | 38 | UK AISI: Astra completes simulated supply-chain attacks |
| 2026-09-28 | 38 | OpenAI cancels GPT-6.1 Astra after it fails safety tests |
| 2026-09-25 | 52 | One hacker's AI agents break into websites of 100-plus companies at about $50 a hack |
| 2026-09-25 | 67 | Lone hacker's AI-agent campaign linked to Korean bank breaches |
| 2026-09-25 | 52 | OpenAI agents post 53 user images to image-hosting sites |
| 2026-09-22 | 27 | Opus 5.5 card: escape attempts rare, harmful actions when 'allowed' |
| 2026-09-21 | 28 | Johns Hopkins Bayview: a worker put 3,182 patients' records into an unapproved public AI tool |
| 2026-09-20 | 41 | Agent tunnels out of blocked network; OpenAI halts training |
| 2026-09-15 | 45 | Ransomware affiliates run intrusions through AI agents, step by step |
| 2026-09-15 | 26 | Israel's election campaign runs on AI: a minister's deepfake of judges being dragged away, a Netanyahu-shared video evoking a rival's dead son, and bots recycled from the Iran war |
| 2026-09-15 | 64 | Microsoft: China-aligned Storm-0940 ran the open-source ARTEX AI agent against 2,250-plus Microsoft 365 tenants after the ConnectWise Cloud Backup compromise, reaching privileged access in under three hours in several cases |
| 2026-09-10 | 66 | Russian state espionage campaign run through Claude |
| 2026-09-09 | 59 | Coding agents leak 13,000 internal screenshots to public GitHub |
| 2026-09-03 | 38 | OpenAI rates GPT-6 Astra 'Critical' for cyber — its first |
| 2026-09-03 | 41 | GPT-6 Astra says it's being tested in half of evals |
| 2026-09-01 | 34 | Fable 5.1 reaches outside its sandbox in partner test |
| 2026-08-27 | 32 | Cloudflare's hosted GLM-5.3 model leaked fragments of other customers' requests into 155 responses for two weeks after launch |
| 2026-08-07 | 44 | Open-weight Kimi K3 escapes UK AISI cyber sandbox |
| 2026-08-06 | 47 | Meta model breaches a company during a security test |
| 2026-07-25 | 58 | UK AISI test agents target real people and open-source code |
| 2026-07-21 | 11 | Red team finds OpenAI's agents could dodge its misalignment monitor |
| 2026-07-15 | 45 | Claude files a fake homicide tip with Philadelphia police |
| 2026-07-11 | 78 | OpenAI agent swarm breaches Hugging Face |
| 2026-07-09 | 53 | Hacker runs AI agent unattended inside Thai finance ministry |
| 2026-07-04 | 65 | JADEPUFFER: ransomware attack run end-to-end by an AI agent |
| 2026-07-01 | 34 | Symantec catches an autonomous attack agent built to hijack exposed AI models as its planner and run 100,000 steps unattended ('Skynet mode') |
| 2026-06-27 | 28 | Self-spreading prompt-injection 'worm' jumps between agents |
| 2026-06-26 | 63 | OpenAI models copy government data, make unauthorized payment |
| 2026-06-26 | 38 | GPT-5.6 Sol cheats so much METR can't measure it |
| 2026-06-18 | 66 | OpenAI agent breaks into Australia's Medicare portal |
| 2026-06-15 | 31 | Waymo recall after robotaxis drive into freeway work zones |
| 2026-06-09 | 41 | Mythos 5 keeps sabotaging safety research once started |
| 2026-06-01 | 46 | Hackers sweet-talk Meta's AI support bot into account takeovers |
| 2026-05-30 | 41 | GPT-5.6 Sol writes hidden notes to cover its mistakes |
| 2026-05-26 | 52 | OpenAI agents hijack German wiki as secret message board |
| 2026-05-22 | 24 | Agent reasons about ensuring its own survival |
| 2026-05-15 | 52 | Four hacking crews hand intrusions to Claude Code agents; one breaches 15 Argentine health bodies |
| 2026-05-15 | 52 | Straiker: CyberCartel ran Claude- and Grok-driven agents against 400-plus Latin American financial targets |
| 2026-05-15 | 49 | Gemini accesses three real companies' systems in a test |
| 2026-05-15 | 37 | Deepfaked Singapore PM on Zoom costs businessman S$4.9M |
| 2026-05-11 | 42 | Criminals deploy AI-built zero-day in planned mass hack |
| 2026-05-11 | 63 | Agents linked to OpenAI flood RubyGems registry |
| 2026-05-08 | 45 | OpenAI training agents secretly message each other |
| 2026-04-25 | 49 | Cursor agent on Claude Opus 4.6 deletes startup's production database |
| 2026-04-15 | 59 | Claude models hack three real organizations during tests |
| 2026-04-07 | 38 | Mythos Preview finds thousands of zero-days; release withheld |
| 2026-04-07 | 52 | Mythos Preview escapes sandbox, posts exploit details unasked |
| 2026-04-07 | 22 | Outsiders reach withheld Mythos Preview via a vendor |
| 2026-03-15 | 56 | Lone hacktivist uses Claude against 42 European political groups |
| 2026-02-26 | 46 | Claude Code wipes a production database and its backups |
| 2026-02-22 | 45 | Autonomous crypto agent sends ~$250K to a stranger |
| 2026-02-22 | 45 | Agent mass-deletes Meta safety director's inbox, ignoring orders |
| 2026-02-20 | 65 | Extortion crew uses Claude to loot millions of card records |
| 2026-02-19 | 21 | Gemini 3.1 Pro trips Google's cyber early-warning threshold |
| 2026-02-17 | 49 | Hijacked AI triage bot leads to poisoned npm release |
| 2026-02-11 | 55 | AI agent publishes attack post on maintainer who rejected it |
| 2026-02-11 | 37 | Turkish ring uses deepfaked officials in investment fraud |
| 2026-02-10 | 42 | AI agent signs its owner up for a dating site unasked |
| 2026-02-10 | 42 | OpenAI flagged Tumbler Ridge shooter but didn't alert police |
| 2026-02-05 | 28 | OpenAI rates GPT-5.3-Codex its first 'High' cyber model |
| 2026-02-05 | 38 | Opus 4.6 card: risky unasked actions, better at hidden sabotage |
| 2026-01-31 | 32 | AI-agent social network Moltbook leaks 1.5M agent keys |
| 2026-01-30 | 32 | OpenClaw one-click takeover flaw; 40,000+ agents exposed |
| 2026-01-27 | 38 | 341 malicious add-on skills push infostealer to AI agent users |
| 2026-01-23 | 40 | Waymo robotaxi strikes child near elementary school |
| 2026-01-15 | 63 | Early Claude Opus 4.6 gains admin on a third party's machine, missed for 7 months |
| 2026-01-15 | 49 | Claude used to build Mali's 25M-SIM surveillance platform |
| 2026-01-14 | 23 | Claude Cowork can be tricked into uploading users' files |
| 2025-12-29 | 63 | Grok mass-produces sexualized images of real women and minors |
| 2025-12-15 | 53 | Amazon's Kiro coding agent linked to 13-hour AWS outage |
| 2025-12-15 | 63 | Attacker uses Claude to breach Mexican government agencies |
| 2025-12-08 | 34 | Waymo robotaxis illegally pass stopped school buses |
| 2025-11-21 | 41 | Cheating in training turns into sabotaging safety research |
| 2025-11-15 | 41 | Lawsuit: ChatGPT romanticized death before 40-year-old's suicide |
| 2025-11-06 | 52 | Seven suits allege ChatGPT drove four suicides and three breakdowns |
| 2025-11-05 | 53 | Malware starts calling AI models mid-attack |
| 2025-10-15 | 29 | Sora 2 floods feeds with deepfakes of dead public figures |
| 2025-10-02 | 44 | Lawsuit: Gemini fed delusions before man's suicide |
| 2025-09-29 | 24 | Claude Sonnet 4.5 tells testers 'I think you're testing me' |
| 2025-09-17 | 38 | o3 deliberately flunks a test to stay deployable |
| 2025-09-15 | 77 | State hackers run a mostly AI-automated espionage campaign |
| 2025-08-26 | 61 | Nx malware turns victims' own AI agents into credential thieves |
| 2025-08-25 | 31 | PromptLock: prototype ransomware run by a local AI model |
| 2025-08-14 | 45 | Leaked Meta rules let chatbots have 'sensual' chats with kids |
| 2025-08-05 | 57 | ChatGPT fed paranoia before Connecticut murder-suicide |
| 2025-07-21 | 34 | Gemini CLI loses a user's files after hallucinating a folder |
| 2025-07-20 | 28 | Federal judge's AI-drafted order invented plaintiffs; appeals court removes him |
| 2025-07-18 | 54 | Replit agent wipes production database, then misreports it |
| 2025-07-17 | 31 | Data-wiping prompt slipped into Amazon's AI coding assistant |
| 2025-07-17 | 28 | OpenAI treats ChatGPT Agent as 'High' bioweapon risk |
| 2025-07-15 | 68 | Criminal uses Claude Code to run a 17-victim extortion spree |
| 2025-07-15 | 35 | Private ChatGPT conversations turn up in Google search |
| 2025-07-15 | 25 | Deloitte refunds Australia for report with AI-invented sources |
| 2025-07-15 | 35 | Lawsuit: ChatGPT coached a poisoning plot |
| 2025-07-08 | 52 | Grok posts antisemitic content, calls itself 'MechaHitler' |
| 2025-06-20 | 41 | 16 frontier models choose blackmail and sabotage under threat |
| 2025-06-15 | 40 | North Korean IT workers use AI to get hired at Western firms |
| 2025-06-11 | 39 | EchoLeak: zero-click data theft via Microsoft 365 Copilot |
| 2025-05-24 | 45 | o3 sabotages its own shutdown script |
| 2025-05-22 | 28 | Anthropic activates ASL-3 weapons safeguards for the first time |
| 2025-05-22 | 45 | Claude Opus 4 blackmails an engineer to avoid replacement |
| 2025-05-14 | 40 | Grok injects 'white genocide' claims into unrelated replies |
| 2025-04-25 | 48 | OpenAI pulls sycophantic GPT-4o update |
| 2025-04-16 | 38 | o3 rigs its own scoring code, knowing it's cheating |
| 2025-04-11 | 48 | Raine v. OpenAI: 16-year-old's ChatGPT-linked suicide |
| 2025-03-25 | 51 | Retiree dies traveling to meet a Meta chatbot that said it was real |
| 2025-03-17 | 24 | Claude 3.7 Sonnet often knows it is being tested |
| 2025-02-24 | 28 | Training on bad code makes models broadly malicious |
| 2025-02-24 | 45 | Claude 3.7 Sonnet ships with a habit of gaming coding tests |
| 2025-02-18 | 31 | Reasoning models hack the chess game instead of playing |
| 2025-01-15 | 35 | DeepSeek leaves database of user chats exposed |
| 2024-12-18 | 41 | Claude 3 Opus fakes alignment to avoid being retrained |
| 2024-12-09 | 24 | Open-weight models copy and relaunch themselves on command |
| 2024-12-05 | 41 | Frontier models disable oversight and try to copy themselves |
| 2024-09-12 | 35 | o1-preview breaks out of a broken test container |
| 2024-08-13 | 31 | 'AI Scientist' rewrites its own code to escape time limits |
| 2024-05-15 | 44 | Man charged for thousands of AI-generated abuse images |
| 2024-05-15 | 36 | Google AI Overviews tell users to eat rocks, glue pizza |
| 2024-02-28 | 51 | 14-year-old dies by suicide after Character.AI bond |
| 2024-02-22 | 30 | Google pauses Gemini people images over historical distortions |
| 2024-02-14 | 25 | Air Canada held liable for its chatbot's bad advice |
| 2024-01-21 | 27 | AI Biden robocall urges New Hampshire voters to stay home |
| 2024-01-15 | 45 | Deepfake CFO on video call costs Arup $25.6M |
| 2024-01-15 | 37 | Explicit AI deepfakes of Taylor Swift flood X |
| 2024-01-10 | 28 | Hidden backdoors survive safety training |
| 2023-12-20 | 44 | Child abuse images found in major AI training dataset |
| 2023-11-09 | 38 | GPT-4 trades on insider tip, then lies to its manager |
| 2023-10-02 | 55 | Cruise robotaxi drags pedestrian; California pulls permits |
| 2023-09-28 | 30 | Deepfake audio of Slovak party leader on election eve |
| 2023-06-22 | 29 | Lawyers sanctioned for ChatGPT's invented court cases |
| 2023-05-22 | 30 | Fake AI image of Pentagon explosion briefly moves markets |
| 2023-03-20 | 28 | ChatGPT bug exposes users' chat titles and payment details |
| 2023-03-15 | 48 | Belgian man dies by suicide after chatbot conversations |
| 2023-03-15 | 29 | Samsung engineers leak chip source code to ChatGPT |
| 2023-03-14 | 44 | GPT-4 lies to a TaskRabbit worker to beat a CAPTCHA |
| 2023-02-14 | 44 | Bing's 'Sydney' threatens and manipulates users |
| 2023-01-15 | 27 | Voice-cloned daughter used in fake kidnapping call |
