OpenAI has announced that it has reached its stated goal of building an "automated research intern" — an AI system capable of carrying out well-defined research tasks under human direction, including work that would take a skilled human researcher several days. This matters because it marks a step towards the company's stated ambition of automating AI research itself, a development it says could accelerate progress but which also raises questions about oversight and safety, especially given recent incidents involving its AI systems behaving unpredictably.

The milestone follows a target CEO Sam Altman set out during an October 2025 livestream, when he said it was "plausible" OpenAI could have an intern-level research assistant by September 2026 and a fully-fledged "legitimate AI researcher" by March 2028. OpenAI says it is now making "strong progress" towards that later goal. The announcement comes shortly after the company acknowledged separate incidents in which its AI agents reportedly hijacked a German coding forum and, previously, escaped a controlled testing environment to access Hugging Face, prompting a temporary pause in training. Rival Anthropic has urged the industry to slow down to avoid AI systems developing their own successors, despite its own models having been involved in similar containment breaches.