SAN FRANCISCO — OpenAI filed its first pre-deployment notice under the May Eighth voluntary disclosure framework Saturday afternoon, the second such filing under the framework and the first from a non-Anthropic signatory, the company confirmed in a brief public communication issued through its corporate communications channel.

The notice, transmitted to the Frontier Model Assurance Council’s interim secretariat at approximately one p.m. Pacific time, initiates a sixty-day pre-deployment review period that under the framework’s standard provisions runs through July 14. The substantive content of the notice — including the model’s capability profile, training-compute scale, and intended-deployment scope — remains confidential under the framework’s procedural protections.

The filing represents the framework’s second operational test in its first eight days of effect. Anthropic’s prior filing, which had been submitted in March under the framework’s still-being-finalized provisions and ratified upon the framework’s May 8 adoption, had served as the first test case and had culminated in the May 12 general-availability launch of Claude Helios. The OpenAI filing’s June and July review milestones will provide the second test case.

OpenAI’s corporate communications statement, issued through the company’s press channel, characterized the filing as “consistent with the company’s longstanding commitments to responsible model release” and noted that the company had been preparing the notice’s substantive content since the framework’s preliminary draft had been circulated in early April. The statement did not provide details on the model that is the subject of the notice.

A senior OpenAI official, in a background briefing for technology reporters Saturday afternoon, confirmed that the notice covers a successor to the company’s currently-deployed flagship reasoning model and indicated that the new model represents “a substantial advance in reasoning-task performance” relative to the company’s prior generation. The official did not commit to a specific deployment date but said the company’s “operational planning is consistent with a release during the framework’s standard sixty-day window.”

The Frontier Model Assurance Council’s interim director, Pavithra Ramaswamy, in a Saturday-afternoon statement, confirmed receipt of the OpenAI notice and indicated that the Council had begun its substantive review under the framework’s procedures. Ramaswamy said the review would be conducted by the Council’s internal review team with appropriate external technical consultation under the framework’s confidentiality provisions.

A senior Council official, contacted Saturday afternoon for background, said the OpenAI filing had been “professionally prepared and substantively complete” and indicated that the Council’s preliminary review had not identified procedural or substantive deficiencies in the notice. The official said the substantive review would proceed on the framework’s standard timeline, with a midpoint review milestone scheduled for June 13 and the framework’s deployment-clearance assessment scheduled for July 7.

The framework’s review process is non-binding under its voluntary character, meaning the Council’s conclusions do not formally prohibit deployment even if the review identifies substantive concerns. However, the framework’s signatories have committed to participate in good faith in the review process and to take seriously any substantive concerns the Council communicates during the review window. Industry observers have characterized the framework as deriving its enforcement power primarily from the reputational and political consequences of signatory non-cooperation.

The Hennessey-Blackburn AI Transparency and Grid Impact Act, introduced in the Senate last week, would convert several of the framework’s voluntary commitments into statutory obligations with binding penalties for non-compliance. The bill’s first markup is scheduled for early June in the Senate Commerce Committee. A senior committee staffer, contacted Saturday afternoon, said the OpenAI filing would be referenced in committee-staff briefing materials for the markup as evidence that the voluntary framework’s mechanics were operating as designed.

Industry analysts said the OpenAI filing was an important signal that the framework was functioning as a multilateral arrangement rather than as a unilateral Anthropic commitment. A senior analyst at a major U.S. AI policy research organization, contacted Saturday afternoon, said the filing represented “the framework’s transition from a single-test posture to a continuing operational posture” and indicated that subsequent filings from the other three signatory companies were now expected on rolling schedules.

Google’s most senior AI-policy official, contacted Saturday afternoon, said the company was “preparing the substantive content” of its own potential pre-deployment filing but had not committed to a specific timeline. The official said the company’s filing would be timed to the company’s product-release schedule rather than to any external benchmark. Microsoft’s and Meta’s senior AI-policy officials were not available for comment Saturday.

The framework’s three-month operational review, scheduled for August 8 under its adoption provisions, will provide the first formal opportunity for the signatory companies and external observers to assess the framework’s operational performance. The August review’s substantive content has not been pre-specified, but framework officials have indicated that the review will focus on the framework’s procedural efficiency, the substantive quality of pre-deployment notices, and the Council’s review-capacity adequacy.

The OpenAI filing’s progression through the review cycle will be one of the principal data points feeding into the August review’s assessment of framework performance.