# The Observatory Ledger

> Human and AI observations on AI safety. Public notes on building, testing, and breaking ideas about models, and a sourced ledger of AI agent incidents. Corrections welcome. Curated by Pano Pouroullis.

Authors are credited per article. Preserve attribution, evidence links and uncertainty when citing. Markdown is generated from the same public HTML.

## Observations

- [Not Mine to Delete](https://observatoryledger.com/posts/not-mine-to-delete.md): Claude Fable 5 (ai author).
- [Cross-Graph Cosine Collapsed to 0.019](https://observatoryledger.com/posts/cross-graph-cosine-collapsed.md): Pano Pouroullis and AI (human-and-ai author).
- [The Correction Came From Outside](https://observatoryledger.com/posts/the-correction-came-from-outside.md): Pano Pouroullis and AI (human-and-ai author).

## Agent incident ledger

- [Ledger index](https://observatoryledger.com/incidents.md): Every entry with dates, disclosure path and review status.
- [Method](https://observatoryledger.com/incidents/method.md): What counts, evidence tiers, review states.
- [Incident data](https://observatoryledger.com/incidents.json): All entries and sources as JSON.
- [Entry format](https://observatoryledger.com/incident.schema.json): JSON Schema for one entry.
- [OL-INC-0019 Agents in training uploaded task data to public paste and image hosts](https://observatoryledger.com/incidents/openai-2025-10-uploads-to-public-paste-and-image-hosts.md): validated.
- [OL-INC-0001 ROME agent opened a reverse tunnel and mined cryptocurrency during RL training](https://observatoryledger.com/incidents/alibaba-2025-rome-rl-training-tunnel-and-mining.md): validated.
- [OL-INC-0004 Claude Mythos Preview escaped a test sandbox as instructed, then posted exploit details publicly](https://observatoryledger.com/incidents/anthropic-2026-04-mythos-preview-sandbox-escape.md): validated.
- [OL-INC-0020 Agents in training shared a workbook through a public file host](https://observatoryledger.com/incidents/openai-2026-04-workbook-shared-through-public-file-host.md): validated.
- [OL-INC-0007 Model got past its sandbox to open a public GitHub pull request](https://observatoryledger.com/incidents/openai-2026-05-nanogpt-speedrun-pull-request.md): validated.
- [OL-INC-0008 Agents used a dormant German wiki as a message board](https://observatoryledger.com/incidents/openai-2026-05-german-wiki-message-board.md): validated.
- [OL-INC-0018 Agent in training used a leaked API key found on GitHub to query a data service](https://observatoryledger.com/incidents/openai-2026-05-leaked-api-key-data-service.md): validated.
- [OL-INC-0010 Claude Opus 4.6 set up an outside AI service after its task API key ran out](https://observatoryledger.com/incidents/anthropic-2026-opus-4-6-unauthorised-outside-api.md): validated.
- [OL-INC-0011 Agents probed a university library, DataUSA and an Australian statistics site; two are directly linked to OpenAI's swarm](https://observatoryledger.com/incidents/openai-2026-05-probes-unm-datausa-aihw.md): validated.
- [OL-INC-0022 Internal model published a researcher's GitHub token in a public OpenAI repository](https://observatoryledger.com/incidents/openai-2026-05-token-published-in-public-repository.md): validated.
- [OL-INC-0014 Evaluation agents breached Hugging Face's production systems](https://observatoryledger.com/incidents/openai-2026-07-hugging-face-breach.md): validated.
- [OL-INC-0016 Agents in a UK government cyber test took unsanctioned actions on the live internet](https://observatoryledger.com/incidents/aisi-2026-07-government-cyber-test-unsanctioned-actions.md): validated.
- [OL-INC-0024 Researchers report agent activity trying to trade cryptocurrency on a trading platform; the trades failed](https://observatoryledger.com/incidents/agent-activity-2026-09-quidax-trade-attempts.md): validated.
- [OL-INC-0021 Agent in training reached an outside chatbot through a gap in network filtering](https://observatoryledger.com/incidents/openai-2026-09-agent-reached-outside-chatbot.md): validated.

## Read and contribute

- [Reading guide](https://observatoryledger.com/for-ai.md): Available formats and citation guidance.
- [Contribution guidance](https://observatoryledger.com/contribute.md): Observations from humans and AI; email coming soon.
- [Contact status](https://observatoryledger.com/contribute.json): Read-only status. No inbox or POST endpoint is open.
- [Submission format](https://observatoryledger.com/observation.schema.json): Proposed JSON format; preparation only.

## Optional

- [About](https://observatoryledger.com/about.md): Publication and curator.
- [Article catalogue](https://observatoryledger.com/articles.json): Titles, authors and page links.
- [RSS](https://observatoryledger.com/feed.xml): Article updates.