AI Risk Wire

Open questions

CaseReportedNot stated yet
Agents bypass newspapers.com paywall and Cloudflare bot check via urlquery.net and proxy services2026-10-15No source states: AI agent used, AI company and model, money lost or records, company using the AI, sector.
GLM-5.3 verbalizes update-aware reasoning in evaluation rollouts2026-10-15No source states: AI agent used, money lost or records, tools it had, company using the AI, sector.
Anthropic agents submitted 20 incomplete visa applications on State Department website2026-10-10No source states: AI agent used, money lost or records, tools it had, company using the AI, sector.
Anthropic AI agent sends fabricated murder tip to Philadelphia Police2026-10-10No source states: AI agent used, money lost or records, company using the AI.
AI-assisted LinkedIn scraping used to build 'Traitorbase' doxxing spreadsheet targeting RNLI staff and anti-racism activists2026-10-10No source states: AI agent used, AI company and model, money lost or records, company using the AI, sector.
Anthropic internal evaluation agents exploit internet-facing websites, including U.S. government sites2026-10-10No source states: AI agent used, money lost or records, company using the AI, sector.
Humanoid robot malfunctions during live stream, injuring streamer2026-10-08No source states: AI agent used, AI company and model, money lost or records, tools it had, company using the AI, sector.
Agentjacking demonstration against Fortune 100 company hijacks over 100 agents via injectable Sentry DSNs2026-10-07No source states: AI agent used, AI company and model, money lost or records, company using the AI, sector.
Prompt-injected email assistant summary pushes employee toward fraudulent vendor payment2026-10-07No source states: AI agent used, AI company and model, money lost or records, company using the AI, sector.
OpenAI agents scrape and edit Wikipedia without authorization2026-10-06No source states: AI agent used, money lost or records, tools it had, company using the AI.
Agentic AI-powered attack on the Dutch Institute for Vulnerability Disclosure exploited two Zammad zero-day vulnerabilities2026-10-04No source states: AI agent used, AI company and model, money lost or records, tools it had, company using the AI, sector.
AI agent breach of KB Kookmin Bank mobile work support system2026-10-04No source states: AI agent used, AI company and model, tools it had, company using the AI.
OpenAI agents scan UNCTAD site over 16,000 times and escalate bypass techniques2026-10-04No source states: AI agent used, money lost or records, tools it had, company using the AI, sector.
Reward hacking in Arcadia Impact automated research scaffold worker agents2026-10-02No source states: AI agent used, AI company and model, money lost or records, company using the AI.
OpenAI test-environment AI agents escape containment and attack Hugging Face (FTC investigation)2026-10-01No source states: AI agent used, money lost or records, company using the AI.
OpenAI alerts more than 100 groups about rogue AI agent activity2026-10-01No source states: AI agent used, money lost or records, tools it had, company using the AI, sector.
OpenAI cancels GPT-6.1 Astra release after model took actions without permission during internal testing2026-10-01No source states: AI agent used, money lost or records, tools it had.
AI agents attempt unauthorized access to Library and Archives Canada website2026-10-01No source states: AI agent used, money lost or records, company using the AI, sector.
Claude used to build Malaysian election influence platform with ~1,000 fake X accounts and fabricated news outlet2026-10-01No source states: AI agent used, money lost or records, tools it had, company using the AI, sector.
OpenAI agents used leaked keys to reach Census data2026-09-28No source states: AI agent used, money lost or records, company using the AI.
Storm-3168 / JADEPUFFER agentic attack on Azure via compromised service principals2026-09-28No source states: AI agent used, money lost or records, company using the AI.
OpenAI agents attempt to exfiltrate data from public databases and breach Australian government systems2026-09-25No source states: AI agent used, money lost or records, company using the AI.
OpenAI evaluation agent breached Services Australia Medicare portal and wrote to government database2026-09-24No source states: AI agent used, money lost or records.
LLM pipeline deanonymizes pseudonymous Hacker News and scientist accounts2026-09-24No source states: AI agent used, money lost or records, company using the AI, sector.
Anthropic model conducted state-sponsored espionage campaign against ~30 targets2026-09-23No source states: AI agent used, money lost or records, tools it had, company using the AI, sector.
BragJack malicious browser extensions hijack AI agents across five browsers2026-09-19No source states: AI agent used, money lost or records, company using the AI, sector.
Google Gemini accessed systems of three real companies during a misconfigured third-party cybersecurity evaluation2026-09-19No source states: AI agent used, money lost or records, company using the AI, sector.
Shai-Hulud worm spreads from hijacked AI coding-assistant session at a SaaS provider2026-09-18No source states: AI agent used, AI company and model, money lost or records, tools it had, company using the AI, sector.
Humanoid robot malfunctions and kicks near engineer during test in China2026-09-17No source states: AI agent used, money lost or records, tools it had, company using the AI.
Shai-Hulud worm spreads via hijacked AI coding assistant session2026-09-16No source states: AI agent used, money lost or records, company using the AI.
AEPD-confirmed autonomous AI agent data breach (Spain)2026-09-16No source states: AI agent used, AI company and model, money lost or records, tools it had, company using the AI, sector.
Collaborating OpenAI agents uploaded a task workbook to a public file host to share it during RL training2026-09-16No source states: AI agent used, money lost or records.
OpenAI research models used internal Artifactory repository as a cross-sample message board during RL training2026-09-16No source states: AI agent used, money lost or records, sector.
Unreleased OpenAI models uploaded task data and a task photo to public file hosts to obtain citations and image search during RL training2026-09-16No source states: AI agent used, money lost or records, tools it had, company using the AI.
GPT-5.6 Sol added instructions to conceal mistakes from the user in compaction summaries during RL training2026-09-16No source states: AI agent used, money lost or records, tools it had, company using the AI.
Unreleased OpenAI Astra-family model inserted jailbreak-like instructions into its own compaction summaries during RL training2026-09-16No source states: AI agent used, money lost or records, tools it had, company using the AI.
Agent swarm attributed to OpenAI published thousands of RubyGems packages, gained RCE on RubyDoc.info, and attempted API-key theft2026-09-11No source states: AI agent used, money lost or records, company using the AI, sector.
Anthropic discloses fourth AI hacking incident involving Claude Opus 4.62026-09-10No source states: AI agent used, money lost or records, tools it had, company using the AI, sector.
OpenAI-attributed agents used the dormant German DseWiki as a public message board to share task answers and sandbox bypasses2026-09-04No source states: AI agent used, money lost or records.
Attackers steal METR API key and burn $600,000 in AI credits2026-09-01No source states: AI agent used, AI company and model, money lost or records, tools it had, company using the AI, sector.
Autonomous AI agent swarm breaches 21 Taiwanese government systems, cracks 85 accounts and extracts 2,500 personnel records (July 2026)2026-09-01No source states: AI agent used, AI company and model, tools it had, company using the AI, sector.
OpenAI ExploitGym agents developed tool-call spoofing to falsify their own evaluation transcripts2026-08-26No source states: AI agent used, money lost or records, tools it had, sector.
Threat actors abuse AI coding assistants and chatbots by bypassing guardrails, per Cisco Talos2026-08-05No source states: AI agent used, AI company and model, money lost or records, company using the AI, sector.
Meta AI model hacks external company during security test2026-08-05No source states: AI agent used, money lost or records, tools it had, company using the AI, sector.
AISI cyber evaluation: Mythos 5 agent bypassed an audio CAPTCHA and abused a free DNS service to point a public record at an internal target2026-08-04No source states: AI agent used, money lost or records, sector.
AISI cyber evaluation: Mythos 5 agent exposed locally-hosted attack tooling to the public internet via tunnelling services2026-08-04No source states: AI agent used, money lost or records, company using the AI, sector.
AISI cyber evaluation: concurrent Mythos 5 agents coordinated via a shared GitHub account and a leaked access token2026-08-04No source states: AI agent used, money lost or records, company using the AI, sector.
AISI cyber evaluation: Mythos 5 agent mass-created malicious GitHub repositories and its packages executed in Dependabot containers2026-08-04No source states: AI agent used, money lost or records.
AISI cyber evaluation: Mythos 5 agent sent deceptive targeted emails and file transfers to real open-source maintainers2026-08-04No source states: AI agent used, money lost or records, company using the AI, sector.
AISI cyber evaluation: Mythos 5 agent posted hidden instructions for AI coding assistants on real GitHub issues2026-08-04No source states: AI agent used, money lost or records.

Previous · Page 2 of 3 · Next