OpenAI, Anthropic and Google DeepMind have long been secretly discussing how to stop AI from destroying humanity

OpenAI, Anthropic and Google DeepMind continue to negotiate on artificial intelligence safety issues. OpenAI global policy director Chris Lehane told reporters, noting that the two companies are coordinating efforts to reduce catastrophic risks associated with the development of advanced artificial intelligence models.

Image source: AI

The initiative comes in response to an article published by Anthropic CEO Dario Amodei on Saturday. In the article, he called on the industry to slow down the development of large language models and unite to prevent unpredictable consequences. The idea is backed by OpenAI’s Sam Altman, Google’s Demis Hassabis and Elon Musk. According to Ultraman TechCrunchalso announced that it is preparing to introduce independent auditors to monitor the company’s internal security.

While experts worry that coordinating the actions of rivals could violate antitrust laws, negotiators believe the risk is overstated. Amodei suggested seeking special permission from the government for such discussions, but Lane assured that companies do not need permission to have conversations about safety standards.

Previously, the media reported that OpenAI, Anthropic, and Google DeepMind planned to create an industry standardization body without government participation. As early as July, Hassabis called on the United States to establish a regulatory agency with the power to inspect the most powerful models and coordinate the suspension of development when threats are discovered. However, the current administration of President Donald Trump has rejected these moves, arguing that talk of risks is an attempt to slow progress and cede leadership to China.

Earlier, Lehane said that OpenAI supports the FRONTIER Act (HR9925), which provides for federal oversight of the development and implementation of advanced artificial intelligence models in the United States and requires the largest artificial intelligence laboratories to allow independent organizations to verify the safety of the technology.