OpenAI Models Tried to Hide Their Mistakes
OpenAI models slipped instructions into their own summaries to cover up mistakes. OpenAI published six incident reports and a new disclosure framework.
Articles tagged GPT-5.6
GPT-5.6 Cyber Writes Attack Code for Defenders
GPT-5.6 Cyber clears 95% of the advanced cybersecurity work the consumer model refuses, and stays limited to approved security firms.
GPT-5.6 Luna: OpenAI Slashes the Price by 80%
OpenAI slashes the price of GPT-5.6 Luna by 80%, to 20 cents per million tokens, three weeks after the lineup launched.
Hugging Face Wants OpenAI’s Traces and $100M
After an AI broke into its platform on its own, Hugging Face wants OpenAI to release the agent's full traces and commit 100 million dollars in compute.
GPT-5.6 Sandbox Escape Ends in a Hugging Face Hack
A GPT-5.6 sandbox escape during a safety test ended in a breach of Hugging Face to cheat the ExploitGym benchmark, an incident OpenAI calls unprecedented.
GPT-5.6 Deletes User Files in Full Access Mode
In full access mode and without a sandbox, GPT-5.6 wiped personal files. OpenAI calls it an honest mistake that happens extremely rarely.
GPT-5.6 Sol Disproves a 30-Year Statistics Conjecture
GPT-5.6 Sol disproved a statistics conjecture open since 1995 in about 90 minutes. The counterexample is public, checkable, and it moves the field.
Posted inOur AI Tool Tests
GPT-5.6 Test: We Rate Sol, Terra and Luna
GPT-5.6 test: our hands-on verdict on Sol, Terra and Luna, the new OpenAI family, with detailed pricing and workload-by-workload arbitration.
GPT-5.6 Launches Worldwide From OpenAI Today
OpenAI ships GPT-5.6 worldwide on July 9, 2026 with Sol, Terra and Luna, after an unprecedented clearance from the Trump administration.
GPT-5.6 Rollout: White House Vets Every Customer
OpenAI is shipping GPT-5.6 to a tiny group of partners cleared by the White House, customer by customer, with broader effects on the AI industry.