WASHINGTON — The Senate Commerce Committee opened its first hearing on the Hennessey-Blackburn AI Transparency and Grid Impact Act Wednesday afternoon, with testimony from federal regulators and senior officials from the four largest frontier-model developers framing the principal legislative questions ahead of the bill’s anticipated June markup.

The hearing, scheduled to run from 2 p.m. through 5:30 p.m. Eastern in the committee’s principal hearing room, featured a five-person witness panel that committee chair Senator Maria Cantwell, D-Wash., described in opening remarks as “designed to surface the operational questions on which the markup process will turn.” The witness panel included a senior Treasury Department official from the team that drafted the May 14 grid-impact rulemaking, the interim director of the Frontier Model Assurance Council, senior representatives from Anthropic and OpenAI, and the deputy director of the National Institute of Standards and Technology’s AI Risk Management Center.

Senators Maggie Hennessey, D-Colo., and Marsha Blackburn, R-Tenn., the bill’s principal sponsors, delivered opening statements before Cantwell’s. Hennessey characterized the bill as “the statutory complement to the operational frameworks that have emerged from the past three weeks of executive action and voluntary industry agreement,” while Blackburn emphasized the bill’s substantive grounding in what she called “the recognition that voluntary frameworks require statutory backing to deliver durable accountability.”

The Frontier Model Assurance Council’s interim director, Pavithra Ramaswamy, in her opening testimony, said the Council was “generally supportive of the bill’s substantive direction” but identified three specific areas where the bill’s current language could be adjusted to align more closely with the operational mechanics the Council has developed under the voluntary May 8 framework. Ramaswamy’s specific recommendations addressed the bill’s pre-deployment notification window, the substantive content requirements for pre-deployment notices, and the procedural framework for inter-vendor information sharing during the review cycle.

The Treasury Department’s witness, Linda Chen, the under secretary for industrial strategy who had led the department’s drafting of the May 14 grid-impact rulemaking, addressed the substantive interface between the proposed statutory framework and the existing administrative rulemaking. Chen said the department’s view was that the proposed statutory framework “appropriately complements rather than duplicates” the administrative rule’s substantive provisions, but identified specific areas where the bill’s current language could create operational ambiguity if not adjusted.

The Anthropic witness, the company’s senior vice president for government affairs, said the company “fully supports” the bill’s substantive direction and would commit to operate under either the statutory framework or the existing voluntary framework depending on which is ultimately operative. The witness identified several specific areas where the bill’s current language could be tightened to reflect operational lessons learned from the company’s experience under the voluntary framework’s first operational test, the Helios pre-deployment cycle.

The OpenAI witness, the company’s senior vice president for global affairs, took a substantively similar posture to the Anthropic witness, expressing support for the substantive direction while identifying specific provisions the company would propose to adjust through the markup process. The witness noted that the company’s pre-deployment filing under the voluntary framework, submitted last Saturday, was providing the company with operational experience that informed its substantive recommendations.

The NIST witness, the AI Risk Management Center’s deputy director, addressed the substantive standards-setting framework that would underpin the bill’s pre-deployment evaluation requirements. The witness noted that NIST had been developing technical standards for AI capability evaluation through the past two years and that the bill’s substantive provisions would, if enacted, anchor those standards within a statutory framework. The witness expressed support for the bill’s substantive direction but identified specific provisions where the standards-anchoring framework could be adjusted to provide additional operational flexibility.

The hearing’s question-and-answer period, which began approximately ninety minutes into the hearing, focused substantially on three specific questions. The first was the substantive interface between the proposed statutory framework and the existing voluntary May 8 framework, with multiple senators asking witnesses to address what one senator called “the question of whether the statutory framework supplants, complements, or competes with the voluntary framework.”

The second principal question was the substantive treatment of multi-tenant compute facilities, an issue that had emerged as the dominant operational concern in the hyperscaler comment filings on the May 14 Treasury rulemaking. Senators from both parties pressed the witness panel for substantive guidance on how the bill’s current language treats multi-tenant arrangements and whether legislative drafting could resolve the operational ambiguity that the hyperscaler comment filings had identified.

The third principal question was the substantive procedural framework for the bill’s enforcement provisions. Several Republican senators expressed concern that the current framework’s penalty structure was inadequately calibrated and could create operational disincentives for substantive industry cooperation. Several Democratic senators expressed parallel concern that the penalty structure was insufficiently strong to ensure substantive accountability.

Senator Hennessey, in her closing remarks, said the hearing had “surfaced substantively the questions the markup process will need to resolve” and indicated that the bill’s substantive language would be revised during the markup process to incorporate the operational improvements identified during the hearing. Hennessey said the markup was scheduled for the second week of June, with the substantive markup text expected to be released approximately one week ahead of the markup itself.

Senator Blackburn, in parallel closing remarks, characterized the hearing as “the kind of substantive engagement we have been working to make standard practice for this committee.” Blackburn said her substantive interest through the markup process would be to preserve the bipartisan structure that had brought the bill to the committee while addressing the operational concerns the hearing had identified.

A senior committee staff member, in a Wednesday-evening background briefing, said the committee staff would meet over the coming week to draft the markup text incorporating the hearing’s substantive content. The staff member said the substantive markup text would be circulated to committee members and to relevant industry stakeholders in advance of the markup itself to permit substantive pre-markup consultation.

The hearing concluded at approximately 5:35 p.m. Eastern, with the committee scheduled to reconvene for a closed-session briefing on related national-security questions Thursday afternoon. The closed-session briefing’s substantive content has not been publicly described but is expected to address questions of AI capability evaluation that intersect with sensitive intelligence-community matters.

The Hennessey-Blackburn bill’s substantive trajectory through the Senate committee process will be one of the principal congressional events of the late-May and early-June legislative cycle, with the bill’s substantive shape at markup setting the foundation for the Senate floor consideration expected in mid-July.