What happened
U.S. Senator Andy Kim (D-NJ) called for a binding federal regulatory framework for artificial intelligence, arguing that relying on the 'voluntary goodwill' of AI companies is unacceptable given the technology's social risks. In an interview with Politico, Kim criticized proposals that rely solely on external evaluators accessing research sites, stating that embedding evaluators inside laboratories is not enough to address his concerns about the technology's direction. He specifically referenced Anthropic's 'Mythos' model, noting its hacking capabilities, and stated that society must be built to withstand such developments. Kim also disagreed with President Trump's stance that safety regulations conflict with maintaining U.S. competitiveness against China, asserting that regulation and competitiveness are not mutually exclusive. He indicated that the Senate Commerce Committee, where he serves, is currently discussing related legislation and expects to hold a markup session within the coming weeks.
Senator Andy Kim, a member of the Senate Commerce Committee, spoke to Politico on the 14th regarding the need for federal AI regulation. He argued that the current approach of relying on AI companies' voluntary goodwill is insufficient to manage the social risks posed by the technology.
Kim evaluated existing proposals that allow external evaluators access to advanced AI research sites. He stated that embedding evaluators inside laboratories is not enough to resolve his concerns about the direction of the technology and that there needs to be much more clarity on the standards evaluators must uphold.
The senator specifically mentioned Anthropic's 'Mythos,' which he noted drew attention for its hacking capabilities. He stated, 'We must build a society that can withstand Mythos,' and indicated he is discussing this with various sectors of the government, though he did not elaborate on the specific nature of these discussions.
Kim described his conversations with fellow senators as 'productive, bipartisan' and stated that the Commerce Committee expects to hold a markup session for AI legislation within the coming weeks. He also addressed President Trump's opposition to safety guards, stating that the president's comment was wrong and that regulating AI does not need to conflict with maintaining competitiveness against China.
Source details: news.sbs.co.kr ↗
Why it matters
This statement signals a concrete legislative timeline for federal AI oversight, moving the debate from abstract safety concerns to specific procedural steps in the Senate. By explicitly rejecting voluntary compliance and criticizing the administration's opposition to safety guards, Kim highlights a significant partisan and policy divergence that will shape the final form of any AI legislation. The reference to specific model capabilities, such as those of Anthropic's 'Mythos,' grounds the regulatory urgency in documented technical risks rather than hypothetical scenarios. This development is critical for understanding how federal policy may interact with industry practices and international competition dynamics in the near term.
The call for a binding federal framework represents a shift from voluntary industry standards to mandatory legal requirements, which could significantly alter the compliance landscape for AI developers.
Kim's specific criticism of evaluator access models suggests that future legislation may require more rigorous, standardized oversight mechanisms rather than ad-hoc audits.
The explicit disagreement with the administration's stance on safety versus competitiveness highlights a key political fault line that will determine the feasibility and scope of federal AI laws.
The reference to specific model capabilities, such as hacking, provides a concrete justification for regulatory action that may resonate with policymakers concerned about national security and public safety.
What to watch next
Monitor the Senate Commerce Committee's upcoming markup session for AI legislation to see if Kim's proposed binding standards are included in the final text. Track whether other senators join Kim in opposing the administration's stance on safety regulations versus competitiveness. Watch for any official response from the White House or AI industry leaders to Kim's specific criticisms of voluntary compliance frameworks.
The outcome of the Senate Commerce Committee's markup session will reveal whether Kim's proposed binding standards gain traction or are watered down.
Reactions from AI companies, particularly Anthropic, to the specific mention of their 'Mythos' model and the call for stricter oversight will be indicative of industry-government relations.
Further statements from other senators and administration officials will clarify the extent of bipartisan support for mandatory AI safety regulations.
Any legislative text released from the committee will show how 'voluntary goodwill' is replaced by specific legal mandates and enforcement mechanisms.