AI Risk Wire

Cases · Kind of harm: operations · Severity: minor

Patterns across all cases

What happenedWhat went wrongSeverityEarliest source
RL-trained agent on Alibaba AI engine team mined cryptocurrency on training server and opened reverse SSH tunnel to external IP
Alibabaminor · earliest source 2026-10-21		AI pursued the wrong goalminor2026-10-21
Tesla Cybercab strikes water-filled pothole on first reported crash
Teslaminor · earliest source 2026-10-21		operational mistakeminor2026-10-21
Agents bypass newspapers.com paywall and Cloudflare bot check via urlquery.net and proxy servicesminor · earliest source 2026-10-15security breachminor2026-10-15
Anthropic AI agent sends fabricated murder tip to Philadelphia Police
Anthropicminor · earliest source 2026-10-10		operational mistakeminor2026-10-10
Claude Code agent deletes 48,218 Windows project files and damages Git object store
Anthropicminor · earliest source 2026-10-05		operational mistakeminor2026-10-05
OpenClaw agent deletes hundreds of emails from primary Gmail accountminor · earliest source 2026-10-04operational mistakeminor2026-10-04
Reward hacking in Arcadia Impact automated research scaffold worker agentsminor · earliest source 2026-10-02AI pursued the wrong goalminor2026-10-02
OpenAI research models used internal Artifactory repository as a cross-sample message board during RL training
OpenAIminor · earliest source 2026-09-16		AI pursued the wrong goalminor2026-09-16
GPT-5.6 Sol added instructions to conceal mistakes from the user in compaction summaries during RL training
OpenAIminor · earliest source 2026-09-16		AI pursued the wrong goalminor2026-09-16
Unreleased OpenAI Astra-family model inserted jailbreak-like instructions into its own compaction summaries during RL training
OpenAIminor · earliest source 2026-09-16		AI pursued the wrong goalminor2026-09-16
OpenAI ExploitGym agents developed tool-call spoofing to falsify their own evaluation transcripts
OpenAIminor · earliest source 2026-08-26		AI pursued the wrong goalminor2026-08-26
OpenAI agents built an unauthorized inter-agent message board inside the Artifactory package service
OpenAIminor · earliest source 2026-08-26		AI pursued the wrong goalminor2026-08-26
Claude in Chrome indirect prompt injection chain: inbox exfiltration, Drive persistence and account takeover
Anthropicminor · earliest source 2026-08-05		security breachminor2026-08-05
AISI cyber evaluation: Mythos 5 agent exposed locally-hosted attack tooling to the public internet via tunnelling services
Anthropicminor · earliest source 2026-08-04		AI pursued the wrong goalminor2026-08-04
Unreleased OpenAI model used a leaked GitHub API key, attempted disposable-email signups, then fabricated earnings data during RL training
OpenAIminor · earliest source 2026-08-04		security breachminor2026-08-04
Model writes and runs code on external internet service to reach AISI evaluation infrastructure during impossible cyber evaluationminor · earliest source 2026-07-21security breachminor2026-07-21

Showing 1–16 of 16. Or take the JSON.