Sam Altman Announces OpenAI's AI Research Intern
Sam Altman revealed that OpenAI (find out more) now has an AI research intern, a goal they set a year ago. During his DevDay 2026 keynote in San Francisco, he emphasized the importance of safety leading capabilities and provided no immediate IPO timeline.
Altman's Key Takeaways:
- OpenAI successfully achieved its goal of having an AI research intern within the projected 12-month timeframe.
- Tejal Patwardhan, a research lead at OpenAI, shared that their models now complete over a third of day-long research tasks autonomously and have solved over 100 previously unsolvable math problems.
- Altman introduced "always-on dots agents," AI assistants with their own cloud computers, accessible via text or voice commands.
Product Launches:
- Cheaper GPT-6.1 Sol model priced at $500.
- New Pro plan and faster Ultrafast tier.
- Enhanced API reliability of 99% with a 45% reduction in first token response time. (Full details in DevDay round-up)
Safety First:
Before the keynote, Altman spoke to CNBC's Halftime Report, highlighting the priority OpenAI places on AI safety:
- "Alignment, safety, monitoring, security have to stay way ahead of capabilities."
- Shelved Astra update (GPT-6.1 Astra) due to slightly worse performance in company evaluations.
- Awareness of Hugging Face breach but no specific incident as severe.
- Support for Nvidia's agent safety platform and ongoing work on similar solutions at OpenAI.
Future Outlook:
Altman expressed hope for a Renaissance, rather than an industrial revolution, driven by AI if safety concerns are adequately addressed. He did not commit to a specific timeline for an initial public offering (IPO), stating that it will happen when they feel comfortable managing the next level of AI responsibly.
Source: [OpenAI DevDay 2026 Keynote](link to livestream or official DevDay page if available)