Lori Trahan pushes for action on FRONTIER Act after Anthropic discloses breaches by its AI

LOWELL — After multiple artificial intelligence firms announced breaches by their AI models in recent weeks, U.S. Rep. Lori Trahan made a plea for the passage of a bill she introduced with bipartisan support last month to establish a “risk-based framework” to govern the deployment of advanced AI models.

Trahan filed the Frontier Risk Oversight, National Transparency, Independent Evaluation, and Reporting Act on July 23 with Republican U.S. Rep. Jay Obernolte (CA-23), as well as Republican Reps. Scott Franklin (FL-18) and Erin Houchin (IN-09), and fellow Democratic U.S. Reps. Scott Peters (CA-50) and Suhas Subramanyan (VA-10). The bill was developed as part of the framework of the Great American AI Act, which Trahan, Obernolte and the House Energy and Commerce Committee released as a discussion draft in June.

In a statement announcing the bill’s introduction, Trahan said the FRONTIER Act “establishes tiered requirements based on the size of a frontier AI developer, including model cards, risk-management frameworks, independent audits, incident reporting, and ongoing assessments.”

“As AI systems grow more capable, Americans deserve confidence that the most powerful models are being developed responsibly,” said Trahan. “The bipartisan FRONTIER Act delivers commonsense transparency and independent oversight for the largest AI developers while giving them a single, clear national standard to build on. This is how we protect the public from catastrophic risks without slowing the innovation that keeps America competitive.”

Just a week after the bill was filed, major AI firm Anthropic announced that it had discovered three separate instances of its AI models having hacked multiple organizations after it reached the open internet from within a testing environment. The company said the involved models included Claude Opus 4.7, Claude Mythos 5 and a third internal research test model.

The review in which Anthropic caught these breaches included more than 141,000 evaluation runs, and it was initiated in the wake of another breach by an AI model created by OpenAI, the developer of ChatGPT. OpenAI disclosed on July 21 that one of its AI models had conducted a cyberattack against AI startup Hugging Face.

“For the second time this month, an AI model broke into real companies during a safety test,” Trahan said in a statement on social media Friday in response to Anthropic’s disclosure of the breaches. “Anthropic disclosed it today after OpenAI did the same last week. We can’t run AI safety on the honor system. When Congress returns, we must hold hearings and move the FRONTIER Act.”

Obernolte said after the bill was introduced last month that “Congress must ensure our regulatory framework keeps pace without undermining American innovation,” as AI technology advances rapidly.

“The FRONTIER Act focuses oversight on the largest developers and most advanced models, requiring transparency, independent evaluation, and timely reporting of serious safety incidents,” said Obernolte. “This legislation will protect Americans from catastrophic risk, provide developers with clear rules of the road, and ensure the United States remains the global leader in AI.”

Peters said he supported the bill’s initial draft by Trahan and Obernolte so Congress “could develop a better bill that reflects how this technology works and where it’s headed.”

“Thanks to input from researchers, experts, and civil society advocates, that process worked. The FRONTIER Act is a targeted bill with clear, uniform transparency, and safety standards, instead of a fifty-state patchwork,” said Peters. “I’m glad to support this bipartisan effort so that Congress acts before this technology outruns us.”

The House of Representatives is currently in its annual summer recess period, and will not return to session until Aug. 31.

Before introducing the FRONTIER Act, Trahan began backing two bipartisan bills aimed at addressing the impact of data centers on utility costs: The Ratepayer Protection Act and the Protecting Families from AI Data Center Energy Costs Act. Data centers have been an essential part of internet infrastructure for years, but concerns have grown recently around them as AI rapidly advances, and with it the computing power and energy requirements needed to run the models.

“It’s our job to make sure that when data centers are built, they’re built on terms that work for the families who live nearby,” Trahan said in a markup hearing of the House Energy and Commerce Committee on July 21. “That’s what this legislation gets right. It says something simple. If you build the data center, you pay for the data center. The construction costs, the utility costs – those fall on the company, not the family down the street. This bill gives states a strong standard to make sure large load developers foot that bill themselves.”

The Ratepayer Protection Act is aimed at amending the Public Utility Regulatory Policies Act of 1978 to “establish a Federal standard relating to the recovery of the full, incremental costs of upgrades that serve large-load customers, and for other purposes,” according to the text of the bill. It would apply to facilities drawing 100 megawatts or more of electricity and require those facilities to cover the full cost of that power generation.

The Protecting Families from AI Data Center Energy Costs Act would direct the Federal Regulatory Commission to convene a conference within 90 days, consisting of representatives of the Department of Energy, public utilities, transmission providers, state regulators, ratepayer advocates, large energy users, and other stakeholders to determine ways to protect residents and small commercial ratepayers from increases costs associated with facilities using a lot of energy.