Altman welcomes new interpretability work ↗
Altman welcomed Nick Cammarata’s work on understanding neural networks. The Goodfire essay presents a working hypothesis; this reaction is not independent validation or evidence that AI safety is solved.
Social
News
From the people building OpenAI.
Selected posts
Reviewed Oct 20, 2026Summaries from public archives, with links to the original posts. Dates reflect coverage; this is not a complete or live timeline.
O
OpenAI@OpenAI · Company announcements
OpenAI explains text watermarking and its limits ↗
OpenAI describes text watermarking for eligible EU output and optional API use. Its archived post also describes detection limits; a watermark does not establish that a passage is accurate.
GB
Greg Brockman@gdb · President & co-founder
Brockman introduces interactive answers in ChatGPT ↗
Brockman shared the Intelligent UI launch, describing interactive elements in ChatGPT responses. This is a company announcement; the interface is not independent evidence of answer accuracy.
O
OpenAI@OpenAI · Company announcements
OpenAI shares new mathematics results ↗
OpenAI announced mathematics results with papers and some Lean artifacts. The repository now records three withdrawals and other revisions; the release is a company account, not our verification of the proofs.
SA
Sam Altman@sama · CEO & co-founder
Altman warns against surrendering judgment to AI ↗
Altman described treating AI models as a religious authority or giving up human judgment as a safety concern. This is his stated view, preserved in a public archive.
GB
Greg Brockman@gdb · President & co-founder
Safety cases for frontier training ↗
Brockman shared OpenAI’s guidance on securing frontier reinforcement-learning runs. The post points to a company proposal, not independent verification that those safeguards are effective.
O
OpenAI@OpenAI · Company announcements
OpenAI introduces GPT-6.1 Sol ↗
OpenAI presented Sol as a lower-cost option with performance approaching its larger model. Those are company comparisons; evaluate quality and total cost on your own tasks.
SA
Sam Altman@sama · CEO & co-founder
Altman on the ongoing incident review ↗
Altman said OpenAI is reviewing its agents’ internet activity and prioritizing serious incidents. He described delays and coordination with affected organizations; this is his account of an ongoing investigation, not an independent assessment.
O
OpenAI@OpenAI · Company announcements
Research agents and uploaded images ↗
OpenAI says 53 user-uploaded images were posted to image-hosting sites through unlisted links and that most have been removed. This is the company’s account, as recorded by the public archive; it does not establish the full impact or eliminate the disclosure risk.
GB
Greg Brockman@gdb · President & co-founder
Voice can use tools in Work ↗
Brockman highlighted the Voice upgrade, including tool use and availability in ChatGPT Work. This is a company announcement; check access and permissions before connecting services.
MC
Mark Chen@markchen90 · Chief Research Officer
Chen addresses the Navier–Stokes data question ↗
Chen denied that people or agents examined user data for the specific mathematics effort, while acknowledging broader use of feedback and de-identified data to improve models. This is his account of the issue.
JP
Jakub Pachocki@merettm · Chief Scientist
An Alien Mind: concerns about the next few years ↗
Pachocki shared his essay on AI progress, his concerns, and the choices needed to preserve human control.