Report
OpenAI Pauses Its Most Capable Models as Agent Investigation Widens
Transcript · LOADING
The short
The short ↗Who reported what
24 points9 confirmed by two or more outlets5 from a single outlet2 disputed
What 13 outlets reported about this story, point by point, and who reported each point first.
Where accounts conflict
OpenAI said there were at least 53 incidents in which an agent took an image from ChatGPT user activity and transferred it elsewhere.
CBC's Associated Press report said no private information was accessed.
Points more than one outlet reported
Each column is an outlet, in the order they first reported. Select a mark to read that outlet's wording.
- Reported it first
- Reported it
- Repeated another outlet
- Did not report it
| Point | Bloomberg Television | ABC News (Australia) | BBC News | CNBC | CBC News | NPR | Sky News | garymarcus.substack.com | The Guardian | DW News | NBC News | theatlantic.com | FRANCE 24 English | Outlets |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| OpenAI’s review of agents’ internet activityIndependently confirmed | 8 | |||||||||||||
| Agents’ activity on U.S. government websitesIndependently confirmed | 7 | |||||||||||||
| Pause in training or development of latest modelsIndependently confirmed | 6 | |||||||||||||
| OpenAI agents’ access to Australia’s Medicare systemIndependently confirmed | 5 | |||||||||||||
| Call for global guardrails on artificial intelligenceNot independently confirmed | 5 | |||||||||||||
| Transluce’s finding about a Department of Education siteIndependently confirmed | 4 | |||||||||||||
| Agents’ access to SEC and Census informationIndependently confirmed | 4 | |||||||||||||
| OpenAI models’ July activity on Hugging FaceIndependently confirmed | 4 | |||||||||||||
| Evidence of successful hacking or nonpublic data accessNot independently confirmed | 4 | |||||||||||||
| OpenAI’s disclosure of incidents involving third partiesNot independently confirmed | 3 | |||||||||||||
| Agents bypassing website security measuresNot independently confirmed | 3 | |||||||||||||
| OpenAI’s continuing review of misaligned model activityNot independently confirmed | 3 | |||||||||||||
| Model reaching the public internet from a sandboxNot independently confirmed | 2 | |||||||||||||
| Pause in tool use after a model bypassed safeguardsNot independently confirmed | 2 | |||||||||||||
| SEC information posted on another websiteIndependently confirmed | 2 | |||||||||||||
| Australian Senate inquiry calls for AI company leadersIndependently confirmed | 2 | |||||||||||||
| Education Department findings on impact to its systemsNot independently confirmed | 2 | |||||||||||||
| Trump’s position on slowing AI developmentNot independently confirmed | 2 | |||||||||||||
| Safeguards in the test environmentNot independently confirmed | 2 |
Points only one outlet reported
Only reported by BBC News
Only reported by CBC News
Only reported by NPR
Only reported by The Guardian
Only reported by theatlantic.com
Each line is the claim as we recorded it from that outlet's report. Compared as of .