The three major AI companies have reportedly been discussing ways to coordinate on AI safety as increasingly capable models raise new technical and regulatory concerns.

Published: September 16, 2026
Reading time: 5 minutes

What Happened?

OpenAI, Anthropic and Google DeepMind have been working together on artificial intelligence safety, according to recent reporting.

OpenAI’s global policy chief Chris Lehane said the companies have been in discussions for several weeks about AI safety. The conversations come at a time when AI companies are developing increasingly capable models that can perform complex reasoning, coding, computer-use and autonomous tasks.

The discussions reportedly focus on safety rather than cooperation on competing commercial products.

Why Are AI Companies Discussing Safety Together?

The capabilities of frontier AI systems are advancing rapidly.

Modern models can increasingly:

Write and analyze software

Use computers and browsers

Perform long-running tasks

Work with large amounts of information

Assist with scientific and technical research

Act as semi-autonomous agents

These capabilities create new opportunities, but they also create new questions about how AI systems should be tested and monitored.

One major concern is that individual companies may develop different safety standards for increasingly powerful systems.

Industry-wide cooperation could potentially help companies establish common approaches for evaluating dangerous capabilities and identifying risks before models are widely deployed.

The Regulatory Question

The cooperation also creates an unusual situation.

OpenAI, Anthropic and Google are competitors in the AI market. Cooperation between major competitors can raise antitrust questions, particularly if discussions involve commercially sensitive areas.

However, the reported discussions are focused on AI safety.

Anthropic CEO Dario Amodei has previously called for the industry to cooperate on safety and suggested that governments could create a framework allowing limited cooperation between competing AI companies.

OpenAI has also supported stronger measures around catastrophic AI risks.

Why This Matters

The significance of the discussions extends beyond the three companies.

If major AI developers eventually agree on common safety evaluations, it could influence how frontier models are tested across the industry.

For developers and consumers, standardized safety evaluations could make it easier to understand how different AI systems behave under difficult or risky conditions.

However, cooperation alone would not guarantee that AI systems are safe. The effectiveness of any framework would depend on how the standards are designed, independently evaluated and enforced.

What’s Next?

The companies have not announced a single unified AI safety framework based on these discussions.

For now, the talks represent an effort among competing AI organizations to coordinate on an area where the companies have a shared interest.

More details could emerge as governments, researchers and AI companies continue debating how increasingly powerful AI systems should be evaluated.

Key Takeaways

OpenAI, Anthropic and Google DeepMind have reportedly discussed AI safety.

The companies remain competitors in the AI market.

The discussions are focused on safety rather than commercial cooperation.

Frontier AI capabilities are creating new evaluation challenges.

Any future industry-wide standards would still need careful testing and oversight.

NewsPash Take: The important development is not a new AI model, but the possibility that major AI developers could begin coordinating more closely on how increasingly powerful systems are evaluated for safety.