
Astra’s math gains came despite OpenAI prioritizing RSI and alignment over targeted math optimization.
The EU Commission is classifying ChatGPT as a very large search engine under the Digital Services Act, triggering new obligations for OpenAI by the end of December 2026.

The EU Commission is classifying ChatGPT as a very large search engine because the chatbot can search the web and answer user queries. The classification applies because ChatGPT reaches at least 45 million monthly users in the EU, the threshold cited under the Digital Services Act. The move places ChatGPT under stricter oversight for the first time as a search engine.
OpenAI has four months, until the end of December 2026, to meet additional obligations under the DSA. Those requirements include risk assessments, semiannual transparency reports, a publicly viewable ad archive, reporting systems for illegal content, a complaints process, and a crisis response mechanism. The Commission also gains broader audit powers over ChatGPT’s risks and operations.
The new rules include data access for approved researchers, but the scope of that access remains disputed. Legal experts disagree on whether it could extend to training data or model weights if needed to assess systemic risks or mitigation measures. The classification does not explicitly allow the Commission to test the models directly.
The Commission is also reclassifying Reddit and Roblox as very large online platforms because users can share content from other users there. All three services now face stricter oversight tied to risks such as illegal content, dangers to minors, and election interference. In total, 28 platforms and search engines are now classified under the DSA.

Astra’s math gains came despite OpenAI prioritizing RSI and alignment over targeted math optimization.

OpenAI’s new speech API supports simultaneous listening and speaking for voice apps.
Paul Christiano says stronger AI alignment and global coordination are needed to reduce catastrophic risk.

An Anthropic-linked dispute puts AI extinction risk, alignment, and lab incentives back in focus.