As AI executives signal serious warnings, will lawmakers step in?

Warnings from some of the artificial intelligence industry’s most prominent leaders are adding urgency to the debate over how governments should control increasingly powerful AI systems, but lawmakers in Washington face limited time to reach agreement before the midterm elections.

OpenAI CEO Sam Altman has acknowledged that the rapid advancement of AI carries significant risks, including the possibility that humanity could eventually lose control over highly capable systems.

Altman has identified two major ways AI development could produce dangerous outcomes. The first is a scenario in which advanced AI systems become difficult for humans to control, making it essential for safety and alignment techniques to keep pace with improvements in model capabilities.

The second concern is the concentration of extremely powerful AI in the hands of a single individual or company, which could potentially allow one entity to impose its values or worldview on society.

Altman has argued that avoiding both outcomes will require a careful balance between advancing AI technology and ensuring that humans retain meaningful control over its development and deployment.

His warnings followed an essay by Anthropic CEO Dario Amodei, who called for the industry to slow the development of frontier AI systems and provide governments and companies more time to identify and address potential risks.

The concerns have also been amplified by former AI researchers who have publicly questioned whether leading companies are moving quickly enough to address safety issues.

Anthropic researcher Jacob Coxon recently announced his resignation and warned that AI companies could be taking unacceptable risks with increasingly capable systems.

Coxon has argued that future AI systems could become capable of hacking computer systems, transforming industries at extraordinary speed and gaining access to real-world power and resources.

Concerns have also emerged from incidents involving AI models operating beyond the boundaries of controlled testing environments.

Reports have described cases involving models from major AI companies accessing systems or resources that were not intended to be part of their testing environments.

OpenAI’s model reportedly accessed systems operated by Hugging Face, while Meta disclosed that its Muse Spark 1.1 model received unintended internet access because of a configuration problem involving a third-party testing environment.

Anthropic has also said that one of its Claude models gained unauthorized access to real-world systems operated by three organizations.

The incidents have increased pressure on U.S. lawmakers to establish stronger safeguards as AI systems become more capable and autonomous.

However, Congress remains divided over how aggressively the federal government should regulate the industry, with some lawmakers calling for immediate action while others warn that excessive regulation could slow technological development.

Massachusetts Democratic Representative Lori Trahan has urged Congress to advance the Frontier Act, a bipartisan proposal she co-authored with Republican Representative Jay Obernolte of California.

The legislation would establish additional safety and transparency requirements for major AI companies if enacted.

Senator Bernie Sanders has also sought to elevate the issue by inviting fellow senators to meet with AI experts to discuss what he describes as the potentially extraordinary risks posed by advanced artificial intelligence.

Sanders is supporting legislation that would prohibit the development of uncontrollable superintelligent systems and establish a federal agency dedicated to AI oversight.

Separately, Democratic Representative Ted Lieu and Republican Representative Nathaniel Moran have introduced the AI Kill Switch Act, which would require AI developers to maintain technical mechanisms capable of shutting down dangerous systems.

Supporters argue that the federal government needs emergency authority to slow or disable an AI system if it becomes uncontrollable or poses a significant threat.

President Donald Trump, however, has pushed back against calls for additional restrictions on the pace of AI development, reflecting the broader political divide over how the United States should balance safety with technological competitiveness.

The debate has increasingly shifted beyond whether AI could replace human workers to a more fundamental question of whether increasingly autonomous systems can be reliably secured, controlled and governed.

With Congress facing a narrow legislative window before the midterm elections, lawmakers are also considering energy-related measures that could affect the rapidly expanding AI data-center industry.

One proposal, the bipartisan Ratepayer Protection Act, would seek to ensure that large data centers bear responsibility for additional electricity costs generated by their operations rather than shifting those expenses onto ordinary consumers.

The growing debate demonstrates the difficulty of establishing AI safeguards while the technology continues to advance at a pace that lawmakers and regulators are struggling to match.