Self-imposed standards outside the Washington umbrella, secret talks between OpenAI, Anthropic and Google to curb frontier model risks

Chris Lehane, global head of public policy at OpenAI, disclosed that his company has been engaged in ongoing talks for weeks with its rivals Anthropic and Google DeepMind to coordinate AI safety standards. The remarks were made during a meeting with journalists in Washington, where Lehane is working with U.S. lawmakers to address the catastrophic risks associated with advanced frontier models, confirming earlier reports that revealed direct consultation channels among the major competing labs to explore joint mechanisms for safety assessment and model governance.

The convergence follows an article published by Dario Amodei, chief executive of Anthropic, in which he called on industry leaders to cooperate in slowing the pace of development of super-intelligent AI models to avoid potentially catastrophic scenarios. Amodei’s appeal received broad support from sector heads, including Sam Altman, Demis Hassabis and Elon Musk via SpaceX AI. Altman, in turn, announced OpenAI’s intention to adopt Anthropic’s approach of allowing independent external assessors access to its systems to monitor safety standards directly and continuously.

The three-way alliance is moving to establish an independent industry standards body without waiting for U.S. government support.Media reports indicate that the three companies are coordinating to set up a unified standards authority for the sector, a direction Altman told his team would materialize apart from official Washington backing. This effort aligns with what Demis Hassabis proposed last July when he urged the U.S. administration to create a regulatory body with the authority to examine the world’s most advanced models and impose a collective slowdown in the industry if technical risks intensify.

These consultations raise delicate legal and political concerns; some industry leaders, including Altman, noted that such coordination could expose companies to antitrust liability if the coordination were deemed an agreement to suppress market competition. While Amodei suggested issuing a narrow government exemption to permit this security collaboration, Lehane argued that firms do not need such an exemption. At the same time, Lehane announced OpenAI’s support for a provision in the Frontier Act that would require frontier labs to host independent audit organizations to ensure model development occurs in a reviewable environment.

The coordination meets explicit opposition from the U.S. administration; President Donald Trump dismissed safety concerns as a hoax, opposing stricter legislation on the grounds that any slowdown in research would give China an advantage in the global race. David Sacks, the administration’s AI adviser and an investor in the field, echoed the same view, asserting that worries about existential risks are overstated and do not warrant curbing the pace of development.

An agreement among frontier giants on unified standards will set the cost of access and usage controls in our regional infrastructure.In my reading of what this alliance means for organizations and tech teams in the Gulf, Egypt and the Levant, the emergence of a self-regulatory standards body led by the major labs means that compliance requirements, safety assessments and the availability of APIs for advanced models will no longer be individual commercial decisions for each provider, but will become synchronized policies that could raise operating costs or constrain the rollout of features. If you are a technical manager overseeing computing infrastructure for a bank in Riyadh or Abu Dhabi, or leading a software team developing banking or educational solutions in Cairo or Amman, exclusive reliance on these labs makes you vulnerable to sudden changes in usage and oversight policies. This three-way coordination compels regional tech leaders to accelerate adoption of open-source models and strengthen sovereign infrastructure, to ensure technical decision-making remains independent and business continuity does not hinge on external lab agreements.