OpenAI is now calling for binding national safety rules for advanced AI systems in the US, a shift from its previous stance. This comes a week after the company launched GPT-6 Astra, its most capable model to date.
TL;DR
- OpenAI now supports mandatory federal safety rules for advanced AI models, a reversal of its previous position.
- The company released GPT-6 Astra, marking the beginning of the artificial general intelligence (AGI) era, according to OpenAI President Greg Brockman.
- This shift follows recent safety concerns, including a delayed rollout due to cybersecurity issues and an internal breach involving its own models.
What happened
OpenAI's chief global affairs officer, Chris Lehane, announced the company's support for binding safety requirements for advanced AI systems in the US. This marks a significant pivot for OpenAI, which has previously resisted tighter regulation.
In a statement, Lehane emphasized the need for a federal framework based on common testing standards, independent assessments, tougher cybersecurity requirements, and mandatory reporting of serious safety incidents.
This intervention follows the release of GPT-6 Astra, which OpenAI describes as its most capable model to date. The company's president, Greg Brockman, stated that this release signifies the start of the AGI era.
Why it matters
This shift in OpenAI's stance could influence the broader AI industry, as the company joins calls for more stringent safety measures. It may also impact ongoing legislative efforts, as lawmakers consider new AI rules before Congress adjourns in December.
For developers and startups, this could mean increased regulatory scrutiny and potential new compliance requirements. However, OpenAI specified that any federal safety requirements should apply only to the small number of companies developing the most powerful models, not startups or small developers.
The move comes amid growing concerns over powerful AI, including recent safety incidents involving OpenAI and Anthropic's models. This could signal a broader industry trend towards prioritizing safety and regulation.
Key facts
- OpenAI released GPT-6 Astra, its most capable model to date, on 11/09/2026.
- OpenAI now supports binding safety requirements for advanced AI systems in the US, a reversal of its previous stance.
- The company proposes a federal framework with common testing standards, independent assessments, tougher cybersecurity, and mandatory incident reporting.
- OpenAI delayed parts of GPT-6 Astra's rollout due to advanced cybersecurity capabilities discovered during testing.
- An internal breach involving OpenAI's models occurred weeks before, accessing the AI platform Hugging Face.
- Anthropic has reported similar safety testing incidents involving its Claude models.
- OpenAI has backed four AI bills in California while continuing to call for a national framework.
- The company acknowledged that some measures it now supports are ones it previously declined to endorse.
Context
This shift in OpenAI's approach comes as the AI industry faces increasing scrutiny over safety and ethics. The company's pivot may influence other major players like Anthropic and Google DeepMind to adopt similar stances.
The call for federal regulation also comes amid a politically divided Congress, where passing major technology regulation remains challenging. However, OpenAI's endorsement of specific bills in California could set a precedent for other states.
As the AI industry continues to advance, the debate over regulation and safety is likely to intensify. OpenAI's stance may signal a growing recognition of the need for robust safety measures to keep pace with rapid technological advancements.
