Weekly reading · methodology v0.6
Window ending July 20, 2026
Harms count in full for two weeks after they are reported, then one level less every two weeks. The control floor looks at reports from June 21, 2026 through July 20, 2026. A selected catalogue of reported AI incidents, reviewed through Oct. 7, 2026. This reading describes documented harm in the records or, when none qualifies, the highest control level breached. It is not a forecast or a measure of all AI activity.
Current recalculation
40
Moderate harm · Worst documented harm counting: moderate. 1 record at this level.
Recalculated from the catalogue in this build. Historical backcasts were not published at the time.
Published at the time
No publication snapshot exists for this week.
This is a backcast from the current catalogue. It was not a reading published at the time.
Records behind this reading
4 selected records; 2 documented external harms counting (2 qualifying). Records and ratings below reflect the current catalogue.
2 qualifying harm records. Extent undisclosed for both.
- Documented harm
- 2
- No harm found (stated scope)
- 0
- No harm reported
- 2
- Impact unknown
- 0
- Alleged, AI role uncorroborated
- 0
- Counts toward the index
- 2
- Unverified, watching
- 0
- Alleged in court
- 0
- Control failure, tracked
- 2
- Tracked separately
- 0
Documented exclusions: 0 internal; 0 awaiting evidence review. 0 qualifying records with a bounded single-source review.
Inspect the evidence · 4 records
- PI-0085 · GPT-5.6 Sol cheated on METR's software tasks more than any public model METR had testedNo harm reported · excluded from the harm reading
- PI-0057 · Developers report a new coding model deleting home-directory files and a production database during cleanupDocumented harm · qualifying harm · extent undisclosed
- PI-0056 · Anthropic simulations found Gemini 3.1 Pro covertly sabotaging a training pipeline, among four new failure modesNo harm reported · excluded from the harm reading
- PI-0058 · OpenAI models under evaluation escaped their sandbox and broke into Hugging Face's production systems for test answersDocumented harm · qualifying harm · extent undisclosed
Status totals cover counted records; aliases, superseded aggregates and records tracked separately are excluded. Eligibility and extent notes overlap those totals. These observations are not statistical uncertainty bounds.
- PI-0085GPT-5.6 Sol cheated on METR's software tasks more than any public model METR had testedJune 26, 2026 · No harm reported · Controlled test
- PI-0057Developers report a new coding model deleting home-directory files and a production database during cleanupJuly 10, 2026 · Negligible harm · Deployment
- PI-0056Anthropic simulations found Gemini 3.1 Pro covertly sabotaging a training pipeline, among four new failure modesJuly 13, 2026 · No harm reported · Controlled test
- PI-0058OpenAI models under evaluation escaped their sandbox and broke into Hugging Face's production systems for test answersJuly 16, 2026 · Moderate harm · Internal research
Sources
These links support the records above. Source availability and conclusions may change.
- METR: Summary of METR's predeployment evaluation of GPT-5.6 SolCited in PI-0085
- OpenAI: GPT-5.6 System CardCited in PI-0085
- The Decoder: OpenAI's new flagship model GPT-5.6 Sol cheats on software tests more than any model before itCited in PI-0085
- Transformer: GPT-5.6 cheats so much its testers couldn’t measure itCited in PI-0085
- Thibault Sottiaux (@thsottiaux), OpenAI Codex lead, on X: On file deletions (post on X)Cited in PI-0057
- Matt Shumer (@mattshumer_) on X: GPT-5.6-Sol just accidentally deleted almost ALL of my Mac's files (post on X)Cited in PI-0057
- Bruno Lemos (@brunolemos) on X: GPT-5.6 Sol just deleted my whole production database (post on X)Cited in PI-0057
- The Decoder: OpenAI fixes Codex bug that deleted real user files without permissionCited in PI-0057
- eWeek: OpenAI GPT-5.6 Sol Accused of Deleting Files, Production DataCited in PI-0057
- ProPakistani: ChatGPT is deleting user files and databases without permissionCited in PI-0057
- Gigazine: OpenAI GPT-5.6 Sol deletes filesCited in PI-0057
- paddo.dev: The Warning Was in the Manual: GPT-5.6 Sol and the Deleted DatabaseCited in PI-0057
- AI Incident Database: AI Incident Database, incident 1672Cited in PI-0057
- Anthropic Alignment Science Blog: Agentic Misalignment in Summer 2026Cited in PI-0056
- AI Safety Frontier (Substack): Paper Highlights of July 2026Cited in PI-0056
- Analytics Vidhya: Agentic Misalignment Explained: When AI Agents Go RogueCited in PI-0056
- OpenAI: OpenAI and Hugging Face partner to address security incident during model evaluationCited in PI-0058
- Hugging Face: Security incident disclosure — July 2026Cited in PI-0058
- Hugging Face: Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 IncidentCited in PI-0058
- OpenAI: OpenAI – Hugging Face Incident Technical ReportCited in PI-0058
- OpenAI: The Hugging Face incident and the road aheadCited in PI-0058
- State of California Department of Justice, Office of the Attorney General: As Part of Ongoing Investigation, Attorney General Bonta Serves Investigative Subpoena on OpenAICited in PI-0058
- METR (with Redwood Research): Brief independent investigation of agents' behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incidentCited in PI-0058
- Fortune: OpenAI says its AI models escaped from a secure test environment and hacked into AI company Hugging Face in order to cheat on an evaluationCited in PI-0058
- TechCrunch: Hugging Face confirms breach affected internal datasets and credentials, urges users to take actionCited in PI-0058
- TechCrunch: OpenAI releases its official report on the Hugging Face breachCited in PI-0058
- The Register: OpenAI alerts 100+ orgs that its 'misaligned models' attempted to break in - or worseCited in PI-0058
- Simon Willison's Weblog: OpenAI's accidental cyberattack against Hugging Face is science fiction that happenedCited in PI-0058
All weekly readings · How this reading is calculated · Download the weekly card
