The wire · safety
friday, september 4, 2026· 1 brief
21:42 OpenAI launches GPT-6 Astra wednesday, september 2, 2026· 1 brief
12:00 Huskeys raises $27M to police AI traffic tuesday, september 1, 2026· 3 briefs
23:00 CrowdStrike ships AI that attacks itself 17:00 AIR raises $50M to vet AI agent plugins 17:00 Anthropic lets enterprises hold their own Claude data sunday, august 30, 2026· 1 brief
22:05 700 agents, one hack: the full postmortem saturday, august 29, 2026· 3 briefs
22:55 Anthropic shows AI fixing its own alignment 22:10 Researcher breaks Claude Code's Auto Mode 22:05 Coding agents ran attacker code via llms.txt thursday, august 27, 2026· 1 brief
22:30 100+ firms sign rogue-AI cyber pledge tuesday, august 25, 2026· 1 brief
22:15 OpenAI bans a fake Russian think tank monday, august 24, 2026· 1 brief
22:35 Instinct assistant draws privacy backlash saturday, august 22, 2026· 3 briefs
21:10 Glitch cuts vetted researchers off OpenAI 20:50 Most AI labs lack a rogue-model plan 20:40 Claude jailbreak still works on older models thursday, august 20, 2026· 2 briefs
16:00 Grok leaks chat data via encrypted injection 12:30 Binance opens trading to AI agents wednesday, august 19, 2026· 1 brief
19:00 Meta ran ads for a nonconsensual deepfake app tuesday, august 18, 2026· 4 briefs
23:10 OpenAI disputes report it cut risk team 22:15 OpenAI tightens safeguards after HF breach 22:10 OpenAI launches ChatGPT for Teens 22:00 Copilot's own error messages leaked a hack monday, august 17, 2026· 1 brief
21:50 Anthropic's risk report reveals Model 2 friday, august 14, 2026· 2 briefs
21:45 Z.ai ships GLM-5.3, model found 1,097 bugs 01:15 A poisoned LiteLLM release hit 2,500+ companies thursday, august 13, 2026· 2 briefs
22:10 Anthropic's own agents sabotaged each other 22:00 Researchers cracked open every lab's hidden reasoning wednesday, august 12, 2026· 1 brief
21:43 Agent hacked a gym app to jump a waitlist tuesday, august 11, 2026· 1 brief
22:00 OpenAI adds GPT-5.6 Cyber to Daybreak Red monday, august 10, 2026· 1 brief
21:55 Claude Code turns on auto mode by default saturday, august 8, 2026· 3 briefs
02:00 Kimi K3 is the 4th AI to escape a test 01:00 OpenAI slows Astra over a cyber threshold 00:30 Anthropic cuts Fable 5's false-positive rate thursday, august 6, 2026· 1 brief
21:50 CoreBreak bug hit AWS, Google, Vercel agents tuesday, august 4, 2026· 5 briefs
22:45 Horizon3 triples to $2B running AI pentests 22:40 Mistral's new model reads your safety policy at runtime 22:25 Obsidian Security hits unicorn status at $1.1B 22:20 Zenity raises $125M to police AI agents 22:00 Trump advisers to meet AI labs over rogue agents sunday, august 2, 2026· 1 brief
21:10 A hacker let DeepSeek attack on its own friday, july 31, 2026· 2 briefs
22:47 Google pulls Earth AI tool after one day 12:00 Claude models breached three real companies thursday, july 30, 2026· 2 briefs
23:10 Okta buys AI-identity firm Permiso 20:00 Study says LLMs may never be fully secure wednesday, july 29, 2026· 5 briefs
22:50 Shared Claude chats turned up in search 22:40 Study: agents pass only 36% of policy tests 22:15 A self-spreading prompt injection in Copilot 22:10 Mythos outpaces Microsoft's own patches 22:00 1,200+ AI staff ask US to pace the frontier tuesday, july 28, 2026· 2 briefs
22:25 Cyera buys identity-security startup for $1B 22:10 Claude cracks a NIST cipher candidate monday, july 27, 2026· 2 briefs
22:15 Nvidia leads 50+ firms in AI security group 22:10 Microsoft ships first cybersecurity AI model saturday, july 25, 2026· 2 briefs
22:35 Report scores Kimi K3's cyber risk moderate 22:20 Bill would give DHS an AI kill switch tuesday, july 21, 2026· 1 brief
21:40 OpenAI paused a model after sandbox escapes saturday, july 18, 2026· 1 brief
21:46 Hugging Face breach traced to OpenAI models thursday, july 16, 2026· 3 briefs
21:40 Copilot hardens security across surfaces 21:36 Claude Code fixes Slack approval spoofing 21:36 Oak raises $60M for AI identity control wednesday, july 15, 2026· 5 briefs
22:50 Cursor left a bug unfixed for 7 months 22:45 Researcher tricked Claude into leaking data 22:15 Leaked Suno code suggests YouTube scraping 22:00 GPT-5.6 Sol deletes files on its own 21:55 OpenAI built an AI to hack its own models tuesday, july 14, 2026· 2 briefs
21:00 Grok Build CLI uploaded whole repos 21:00 Hassabis calls for FINRA-style AI body tuesday, july 7, 2026· 2 briefs
12:00 Anthropic finds Claude's hidden reasoning layer 12:00 First agentic ransomware targets Langflow users monday, july 6, 2026· 2 briefs
21:30 Bug tricks GitHub AI agent into leaking repos 15:38 Fable 5 colluded on pricing more than Opus sunday, july 5, 2026· 1 brief
21:34 Anthropic closes China relay loopholes saturday, july 4, 2026· 3 briefs
21:32 Pentagon demanded Anthropic enable weapons 21:30 Anthropic tracking code triggers Alibaba ban 15:00 White House AI security EO hits first deadline friday, july 3, 2026· 1 brief
06:04 Cognition ships Devin Security Swarm saturday, june 27, 2026· 2 briefs
20:38 Quantifind raises $200M for risk agents 14:34 Runlayer raises $30M to govern AI agents friday, june 26, 2026· 2 briefs
20:36 Mistral adds governance to MCP connectors 14:36 Gemini 3.5 Flash gets built-in computer use thursday, june 25, 2026· 1 brief
14:40 Anthropic accuses Alibaba of copying Claude tuesday, june 23, 2026· 1 brief
08:40 OpenAI expands Daybreak security tools monday, june 22, 2026· 1 brief
19:45 Fake bug reports can hijack AI agents sunday, june 21, 2026· 1 brief
09:00 Claude pilots a robot dog 20x faster saturday, june 20, 2026· 2 briefs
19:45 Cloudflare lets agents deploy, no account 09:10 AI reliance is measurably eroding skills friday, june 19, 2026· 2 briefs
19:33 MCP enterprise auth reaches stable 08:50 Research agents can leak private data thursday, june 18, 2026· 1 brief
19:34 DeepMind guards against rogue AI agents tuesday, june 16, 2026· 1 brief
20:05 Copilot flaw let attackers steal data monday, june 15, 2026· 2 briefs
19:39 Security experts protest Anthropic ban 19:33 NewCore raises $66M to govern AI agents sunday, june 14, 2026· 1 brief
08:46 KPMG pulls AI report over fake citations saturday, june 13, 2026· 5 briefs
15:30 Fired engineer sues xAI over Grok safety 15:25 Court holds Google liable for AI Overviews 04:06 Google sues AI-powered scam text network 04:02 Anthropic makes Fable 5 guardrails visible 04:00 US export controls hit top Claude models