OpenAI, Anthropic and Google are collaborating on AI safety measures

OpenAI, Anthropic, and Google DeepMind have entered a collaboration to address AI safety concerns, with discussions spanning weeks and no antitrust waivers required. The partnership, confirmed by OpenAI’s global policy chief Chris Lehane, follows growing public and political pressure to mitigate risks from advanced AI systems. However, the initiative faces scrutiny over its potential to create industry monopolies and the independence of proposed oversight mechanisms.

Collaboration Details and Legislative Push

The collaboration between the three firms centers on developing shared safety protocols and influencing U.S. legislation to govern AI risks. Lehane emphasized that the companies do not require antitrust exemptions to coordinate on safety, citing precedents like the airline industry’s safety standards. “It’s better to try to work together to prioritize safety,” he stated during a Washington briefing, according to reports from The Hindu Business Line and Bloomberg News.

OpenAI has aligned with bipartisan legislative proposals, including a framework from Republican Representative Jay Obernolte and Democrat Lori Trahan, as well as a Senate bill led by Majority Leader John Thune, Republican Ted Cruz, and Democrat Amy Klobuchar. These measures aim to address “catastrophic risks” from AI, such as loss of control over systems or cyberbiological threats. However, the timeline for legislative action remains uncertain, with Congress expected to adjourn before the November elections, leaving major proposals on hold.

Senator Bernie Sanders, a key advocate for stricter AI regulation, plans to introduce legislation banning “superintelligent AI” by next week. His call for immediate action found rare alignment with conservative figure Steve Bannon, who also criticized the federal government’s inaction. Meanwhile, President Donald Trump has dismissed AI safety concerns, labeling them a “hoax” and opposing regulatory measures.

Watchdog Independence and Financial Conflicts

The collaboration has sparked debates over the independence of AI safety watchdogs. Anthropic’s CEO, Dario Amodei, proposed granting independent evaluators access to AI labs to test systems and report safety issues. However, scientist Kevin Bass raised concerns about potential conflicts of interest, citing financial ties between the Model Evaluation and Threat Research (METR) organization and Anthropic. Bass argued that METR’s funding sources, including billionaire Dustin Moskovitz, could compromise its neutrality.

While METR claims to reject direct funding from AI companies, its reliance on effective altruism-linked donors has drawn scrutiny. Bass called for a congressional investigation into these relationships, though his allegations remain unproven. The controversy highlights a broader challenge: ensuring that oversight bodies are not influenced by the industries they regulate. OpenAI’s Lehane defended the initiative, stating that safety coordination has historical precedents and does not constitute “moat digging” by the firms.

Industry and Political Divisions

Not all industry leaders share the collaboration’s goals. Nvidia CEO Jensen Huang argued that market forces, not government mandates, would drive safe AI innovation. “You can definitely have both at the same time,” he said, dismissing the need for “false choices” between speed and safety. This stance contrasts with OpenAI and Anthropic’s calls for pauses in advanced AI development, which have gained traction following Amodei’s 3,800-word essay urging caution.

Political skepticism also persists. Federal Trade Commission Chair Andrew Ferguson warned that AI firms’ safety coordination could “insulate their business from challengers,” framing it as an antitrust risk. Meanwhile, a proposed bill by Senators Jim Banks (R) and Adam Schiff (D) would grant narrow antitrust exemptions for safety-related information sharing, though it remains part of a larger defense bill with uncertain prospects.

The collaboration underscores the urgent need for consensus on AI governance, even as divisions over oversight, competition, and risk management persist. With the next major test likely to come in the form of legislative proposals or public demonstrations of AI capabilities, the coming months will determine whether the industry’s self-regulatory efforts gain traction or face renewed scrutiny.