Bloomberg Television9 min
Rep. Trahan Says Congress Must Keep Pace With AI
Trahan, Kailey Leinz, Joe Mathieu
Interview, Press Conference
- The potential successful IPO of Anthropic, coupled with the "Jensen Wong scandal" regarding NVIDIA executive influence, is viewed by Representative Chu as a catalyst for accelerating both export controls and the passage of the AI regulatory Frontier Act.
- Congressional Democrats and Republicans are demanding hearings to investigate the administration's knowledge of and timing regarding NVIDIA's chip policy influence.
- The bipartisan Frontier Act proposes three specific requirements: mandatory safety and transparency reporting, incident reporting for models that go "rogue," and a government intervention process for models posing catastrophic risks.
- Senator Elizabeth Warren has raised concerns that the Frontier Act's provisions could negatively impact the financial outlook of major public companies like Anthropic, potentially threatening Americans' retirement portfolios.
- Representative Chu argues there is no trade-off between safety and innovation, stating that the current "honor system" of voluntary compliance is unsustainable and that statutory guardrails are required to maintain public trust.
- The Frontier Act mandates that companies publish safety reports and model cards, report "near miss" events such as the Hugging Face incident, and utilize embedded third-party evaluators rather than self-assessment.
- To pass the Frontier Act, Representative Chu notes the need for 60 Senate votes and the President's signature, emphasizing that a bipartisan bill on the President's desk would pressure the administration to sign it.
- Provisions within the Frontier Act mirror voluntary agreements established in the White House Accords, but the legislation aims to convert these voluntary measures into durable statutory rules.
- Representative Chu supports tightening export controls on advanced chips to China, arguing that allowing a competitor access to US compute infrastructure contradicts the need for AI safety guardrails.
- The discussion regarding stricter export controls is anticipated to be incorporated into the National Defense Authorization Act (NDAA) during the current "lame duck" session.
- The administration's stance on AI pacing appears to have shifted following the NVIDIA scandal, with Representative Chu expressing hope for progress in balancing Chinese competition with safety enforcement.
- Recent widespread outages affecting Microsoft, banks, and airlines were not attributed to rogue AI or cyberattacks by the organizations involved, though they were likely linked to vulnerability surfacing from new AI tools.
- FBI Director notes the use of the "R-TEX" tool by hackers in global attacks, though Representative Chu clarifies the primary concern is AI surfacing vulnerabilities that require rapid patching.
- Representative Chu asserts that the frequency of recent outages underscores the necessity of pre-emptive AI guardrails to build resilience in critical infrastructure rather than reacting to failures.
- The representative calls for increased capability funding for oversight agencies and the implementation of "table stakes" common sense guardrails to prevent catastrophic events.