OpenAI and Anthropic Shape Federal AI Review Thresholds in B

Two of the most prominent names in artificial intelligence are quietly collaborating in Washington on a project that could reshape competition across the entire industry. OpenAI and Anthropic, rivals in the race to build the world's most capable AI systems, have been jointly lobbying the Trump administration to establish federal AI review thresholds that would apply across the sector, including to their most direct competitors at Meta and Elon Musk's xAI.

The shared position from both companies is that pre-release review standards for the most powerful AI models should apply industry-wide rather than being limited to labs that already have established relationships with Washington. With the August 1 deadline fast approaching for defining which models qualify as covered frontier systems under Executive Order 14409, the stakes for every major AI developer have never been higher.

The executive order, signed by President Trump on June 2, 2026, set in motion a 60-day process to build a classified benchmarking framework for frontier AI models. It directed Treasury, Defense, and other federal agencies to develop the structure for evaluating whether a model poses national security risks before it reaches the public. Now, with three days left before that deadline, the details of how that framework will operate are being finalized in negotiations that have largely taken place behind closed doors. The August 1 cutoff marks the first time the US government has set a formal timeline for reviewing frontier AI models before public release.

The Scope of Federal AI Review Thresholds

Under the proposal that OpenAI and Anthropic have been advancing, the government would have the authority to access certain frontier models for up to 30 days before their public release. This review period would apply to models that demonstrate specific capabilities in areas such as cybersecurity or other domains that could present national security risks. The two companies have pushed to expand the type of models that would require this review, arguing that the scope should cover systems built by their competitors as well.

Five labs are jointly developing the threshold criteria: OpenAI, Anthropic, Google, Microsoft, and xAI. This group is the core of the frontier AI ecosystem in the United States, and their participation in drafting the rules they will eventually be measured against has drawn scrutiny from critics who question whether the process amounts to self-regulation dressed as public policy.

Some of the rivals targeted by the expanded review scope have pushed back, preferring that the government take a narrower approach that would not subject their models to the same pre-release scrutiny. The tension reveals a strategic calculation: companies that have already invested in safety infrastructure and government relationships stand to benefit from rules that raise the bar for everyone else. This dynamic has led critics to argue that the rulemaking process risks being captured by the very companies it is meant to oversee.

The Power Play Behind the Push

The collaboration between OpenAI and Anthropic is notable precisely because the two companies compete aggressively for customers, funding, and technical leadership. OpenAI operates ChatGPT and develops the GPT series of models, while Anthropic builds the Claude line. Their joint effort on federal AI review thresholds signals that both organizations see a shared interest in establishing a stable regulatory environment that does not reward companies that move faster on safety and governance.

The dynamic also puts pressure on rivals such as Meta, which releases its Llama models under open-weight licenses, and xAI, which has positioned itself as a fast-moving challenger to the established labs. If the federal review thresholds are written broadly enough to capture open-weight releases or models trained with different safety philosophies, companies that have not participated in the rule-drafting process could face significant delays or compliance costs.

The timing of the effort is particularly strategic. On July 28, just days before the August 1 deadline, more than 1,100 employees from the world's leading AI labs published a public statement titled Pacing the Frontier. The signatories, including Anthropic cofounders Jack Clark and Jared Kaplan, OpenAI chief scientist Jakub Pachocki, and Meta chief scientist Shengjia Zhao, called on Washington to support international efforts to deliberately slow the pace of frontier AI development. Both OpenAI and Anthropic endorsed the statement as companies within hours of its publication, reinforcing the message that the industry itself is asking for stronger oversight.

The employee letter and the corporate lobbying effort are not identical agendas, but they complement each other. The letter creates public legitimacy for the idea that AI progress should be deliberately paced, while the closed-door negotiations give that idea concrete regulatory form. A company cannot credibly endorse a call for slower development while simultaneously opposing the rules that would achieve that goal.

The Voluntary Framework Track

Separate from the executive order process, the White House has been racing to finalize voluntary pre-release testing standards with OpenAI, Google, and Anthropic. Reporting from early July indicated that an announcement could come as soon as the first week of August, aligning with the same deadline that governs the executive order review process.

The voluntary framework is being drafted in close consultation with the same labs it is meant to hold accountable, a fact that has raised questions about how rigorous the standards will ultimately be. The standards would establish benchmarks for advanced AI models and create clearer timelines for testing and launch, but without the binding force of law. Companies that agree to the framework would submit their most powerful models for review before release, creating a layer of accountability that currently does not exist.

The White House has involved OpenAI, Google, and Anthropic directly in the negotiations, but has not extended the same level of participation to other developers. This selective engagement has created a two-tier system in which some companies help write the rules while others must comply with them after the fact. The arrangement mirrors the broader power dynamics taking shape in US AI policy, where the companies with the deepest Washington relationships are positioned to shape the regulatory environment to their advantage.

Competing Visions for Federal Preemption

Anthropic has also been pursuing a parallel strategy at the state level, pushing state legislators to adopt AI safety laws while simultaneously arguing that the authority to block unsafe model deployments should rest with the federal government. The company has been working state by state to ratchet up AI rules, a tactic that OpenAI has opposed through a federal preemption strategy that seeks to prevent states from creating a patchwork of conflicting requirements.

The difference in approach reflects the underlying strategic positions of the two companies. Anthropic, which has positioned itself as the more safety-conscious of the two, benefits from a regulatory environment that raises standards across jurisdictions. OpenAI, with its larger user base and broader product surface, prefers a single federal standard that simplifies compliance and prevents the most aggressive state-level proposals from taking effect.

Despite these tactical differences, both companies have converged on the central point: the federal government should play the primary role in reviewing frontier AI models before release. The disagreement is over how much room state-level experimentation should have in the interim.

Why This Matters

The joint effort by OpenAI and Anthropic to shape federal AI review thresholds is a defining moment in the governance of the most consequential technology of the decade. Two fierce competitors have concluded that the stability of a federal framework is worth more than the short-term advantage of operating without rules. The outcome of this process, expected within days, will determine whether the United States establishes a durable model for AI oversight that other nations may follow. For regulators worldwide observing this process, the credibility of future international AI governance may hinge on whether the thresholds are seen as independently enforced or industry-written. For decision-makers evaluating AI investments and partnerships, the signal is clear: the regulatory floor is rising, and the companies that helped build it will have the strongest footing once it settles.

Sources

Anthropic's Responsible Scaling Policy

✔Human Verified

Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.