OpenAI is working with fellow artificial intelligence companies Anthropic and Google DeepMind on efforts to address AI safety concerns, according to a report from Bloomberg News. The discussions come as concerns about the potential risks of increasingly powerful AI systems continue to grow across the technology industry and among policymakers in the United States.
The reported cooperation reflects a notable development in an industry where companies compete intensely over advanced models, products, talent and computing resources. Despite that competition, AI safety has increasingly become an area where major laboratories may have shared interests. OpenAI, Anthropic and Google DeepMind are among the companies developing increasingly capable AI systems, making questions around safety, testing and potential large-scale risks particularly significant.
OpenAI’s global policy chief Chris Lehane said the company has been in discussions with Anthropic and Google for several weeks, according to Bloomberg News. The talks reportedly focus on safety issues rather than commercial cooperation between the companies. Lehane also indicated that OpenAI does not believe the companies would require an antitrust waiver to coordinate on AI safety matters.
The development comes at a time when debates surrounding the long-term risks of artificial intelligence have become more prominent. Researchers, technology executives and policymakers have raised questions about what could happen as AI systems become more capable and are deployed across a wider range of industries and critical activities. These discussions include concerns about misuse, cybersecurity, autonomous systems, misinformation and other potential consequences of increasingly advanced technology.
The issue has also brought competing AI companies into occasional public discussions about common safety standards. Although companies such as OpenAI, Anthropic and Google DeepMind are pursuing different commercial strategies and developing their own AI systems, they face some of the same technical and regulatory questions. Cooperation on safety could therefore provide an opportunity to share approaches in areas such as evaluating advanced models, identifying dangerous capabilities and developing safeguards.
Lehane said during a briefing in Washington, “It’s better to try to work together to prioritize safety,” according to Bloomberg News. His comments highlight the argument that certain safety challenges may require cooperation even among companies that otherwise compete directly in the rapidly developing AI market.
For OpenAI, the discussions also come as the company continues to engage with policymakers in Washington. Lehane was in the capital meeting lawmakers, where he reportedly said OpenAI would support bipartisan legislation designed to reduce the risk of catastrophic outcomes associated with artificial intelligence.
The involvement of lawmakers reflects the increasingly important role of government in determining how advanced AI systems should be developed and deployed. Governments around the world are considering different approaches to AI regulation, including requirements related to transparency, risk assessment, security, accountability and the testing of high-capability systems. In the United States, the debate has included questions about how much responsibility should rest with technology companies and how much should be established through federal legislation.
Finding common ground on AI safety is not necessarily straightforward. Companies developing advanced models may have different definitions of risk, different internal safety procedures and different views about how regulation should work. There are also practical questions surrounding what information can be shared without exposing sensitive technical or commercial details.
Antitrust considerations represent another part of the discussion. OpenAI’s position, as reported by Bloomberg, is that cooperation between the companies on safety matters would not require an antitrust waiver. That distinction is important because collaboration among major technology companies can attract regulatory scrutiny when it involves areas that could affect competition. Safety cooperation, however, can involve a different set of objectives from agreements concerning pricing, products or market strategy.
The reported discussions also illustrate the unusual nature of the AI industry. Companies are simultaneously competitors and participants in a broader technological ecosystem. An advancement made by one laboratory can influence the direction of the entire sector, while a newly identified safety problem can raise concerns for multiple developers at the same time. This creates incentives for information sharing in certain areas, even when companies remain commercially separate.
AI safety itself covers a broad range of issues. At a technical level, researchers work on methods for evaluating whether models behave as intended, resist harmful instructions and remain controllable under difficult circumstances. Developers also study how models might be misused or manipulated and how safeguards can be tested before systems are made widely available. As AI becomes integrated into business, education, software development, research and other areas, questions about reliability and responsible deployment have become increasingly relevant.
