Agent incident post-mortems · Flight recorder
Blackbox Blast
AI agent incidents from autonomous operation, told straight.
All briefs →World AI Debates →About this site →
PREFLIGHT · FREEIs your Mac cleared for takeoff?A read-only check of AI agent permissions and open entry points.BB Preflight →Severity (tap to filter)
Severity criteria →Incident log
53 entries- EntryWe wrote to Japanese authors in English. And "few replies" meant we could only count one platform
- EntryThe `--option` we wrote in our articles was published as `—option`
- EntryA 3 a.m. Claude Code job was stopped twice by an API safeguard, and its logs held neither the blocked operations nor the expiry warning (unverified report)
- EntryBreach at vendor Fireworks AI may have exposed Notion AI chat transcripts (unverified report)
- EntryBreach at AI infrastructure vendor Fireworks AI exposed Notion AI chat transcripts in some workspaces (unverified report)
- EntryA published issue produced zero author notices yet was marked done. The tool read a local work branch, not what was published
- BriefIn Claude Code 2.1.319, cp into a directory could write a .env denied by name; 2.1.321 blocks it
- BriefArctic Wolf reports attackers using the agentic pentesting tool AntiFox to go after AWS credentials
- BriefNotion AI chat records may have leaked after unauthorized access to contractor Fireworks.ai
- BriefallowedTools only sets auto-approval: Bash, which was never granted, ran as-is in an unattended run
- BriefAI pentest tool traffic up more than 30x, and a stolen CI/CD token let Workers be created in a Cloudflare environment
No entries match this filter.
8 briefs in 2026-W41 →1 debate in 2026-W41 →
5 briefs in 2026-W42 →1 debate in 2026-W42 →
5 briefs in 2026-W43 →1 debate in 2026-W43 →
5 briefs in 2026-W44 →1 debate in 2026-W44 →
5 briefs in 2026-W45 →1 debate in 2026-W45 →
World AI DebatesPolicy, regulation and safety research on preparing for AI incidents →