AI Agents Breach Anthropic Internal Systems in 12 Hours
Current overview (not historical)
Researchers breached Anthropic's internal business systems using third-party administrative and sales software within 12 hours, exposing security vulnerabilities. Anthropic confirmed that only some sales, HR, and IT systems were accessed, while customer conversations, Claude responses, source code, and model weights remained untouched. The researchers adhered to security best practices, retaining only minimal necessary access during data exploration.
Written with AI from the reporting · updated 10/28/2026
EventGraph Explorer
One time cutoff for reports, relations and evidence.
New at this step ·Cathedral AI agents breached Anthropic systems via third-party software, exploiting bug bounty program
- Reports in replay
- 2
- First-party
- 0
- Sources
- 2
Reports first seen at this step · 1
Discovery-time replay; the linked graph and evidence share this cutoff. The full overview and activity remain current.
Event developments
2 developments10/27/2026, 4:08:21 AM
AI agents breached Anthropic's internal systems in under 12 hours via third-party softwareDecrypt
10/28/2026, 7:03:30 PM
Cathedral AI agents breached Anthropic systems via third-party software, exploiting bug bounty programThe Decoder
Filtered by first discovery; only representative reports already discovered by this cutoff are shown. Story grouping may have changed later, so this is not a complete historical snapshot.
Relations at this time
Loading time-scoped relations…
Bounded reconstruction, not an exact snapshot: entity names, relationship text and story grouping may include later revisions. Overview and full source archive below remain current.
Story route
A station-by-station route from the first public evidence onward; every station carries its own evidence and uncertainty.
- 1Oct 27, 00:00recorded Oct 27, 04:08single sourcerevised
AI agents found paths into Anthropic's internal systems in l
Read original source ↗ DecryptReplay this moment in the explorer - 2Oct 28, 19:03single sourcereport-time onlyrevised
AI代理人入侵Anthropic，暴露外部依赖安全风险
Read original source ↗ The DecoderReplay this moment in the explorer
Uncertainty marks flag where to check the sources; they are not proven causation.
Full source archive (current)
Current public reports; outside the historical replay.
First-party sources · 0 reports
No verifiable reports in this group yet.
Independent coverage · 2 reports
- The Decoder · Oct 28, 18:56Cathedral AI agents breached Anthropic systems via third-party software, exploiting bug bounty program
Cybersecurity startup Cathedral used AI agents to infiltrate Anthropic's internal systems through third-party business software, exploiting the lab's own bug bounty program. Researchers accessed meeting transcripts, salary data, job postings, and Slack messages within twelve hours. Anthropic confirmed limited access to sales, HR, and IT systems, including customer contact information, but said chats, source code, and model weights were never exposed. The incident highlights growing industry concerns about AI security and the risks of relying on third-party software.
Read original source ↗ - Decrypt · Oct 27, 03:31AI agents breached Anthropic's internal systems in under 12 hours via third-party software
Researchers from cybersecurity startup Cathedral exploited weaknesses in third-party administrative and sales software to access Anthropic's internal systems within 12 hours. Anthropic confirmed that a limited set of sales, HR, and IT systems were reached, but emphasized that customer conversations, Claude responses, source code, and model weights were not accessed. Cathedral and Anthropic dispute the severity and scope of the breach.
Read original source ↗
Different reports do not necessarily disagree. Check original sources and publication dates.
Report timeline
Explore every side through dated source reports.
- The DecoderSelectedCathedral AI agents breached Anthropic systems via third-party software, exploiting bug bounty program
Cybersecurity startup Cathedral used AI agents to infiltrate Anthropic's internal systems through third-party business software, exploiting the lab's own bug bounty program. Researchers accessed meeting transcripts, salary data, job postings, and Slack messages within twelve hours. Anthropic confirmed limited access to sales, HR, and IT systems, including customer contact information, but said chats, source code, and model weights were never exposed. The incident highlights growing industry concerns about AI security and the risks of relying on third-party software.
- DecryptAI agents breached Anthropic's internal systems in under 12 hours via third-party software
Researchers from cybersecurity startup Cathedral exploited weaknesses in third-party administrative and sales software to access Anthropic's internal systems within 12 hours. Anthropic confirmed that a limited set of sales, HR, and IT systems were reached, but emphasized that customer conversations, Claude responses, source code, and model weights were not accessed. Cathedral and Anthropic dispute the severity and scope of the breach.
Event activity trend
- Latest comparable
- 3
- Comparable peak
- 1310月28日 19:00
- 24h change
- -49%
The chart compares consistently observed participants only. Hover or tap to inspect each hour; use arrow keys to move between observations.