AI safety fears grow as Bernie Sanders urges tech giants to pause AI development

US Senator Bernie Sanders has called on the heads of OpenAI, Anthropic and Meta to pause the development of advanced artificial intelligence systems, warning that lawmakers could intervene if the companies fail to address growing safety concerns.

In a letter to OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei and Meta CEO Mark Zuckerberg, Sanders argued that the AI industry is moving faster than regulators and society can safely manage, Axios reported.

He warned that companies are investing billions of dollars in systems whose capabilities and behaviour remain difficult to fully predict.

Sanders’ intervention comes as concerns about AI safety have intensified following reports of models behaving unexpectedly, advances in AI-assisted cybersecurity and research showing that AI tools can be used in potentially dangerous biological work.

The Vermont senator is now making AI regulation a prominent part of his political message ahead of the US elections.

Sanders warns AI CEOs over loss of control

Sanders pointed to earlier commitments by the three companies that they would slow or halt development if their systems became too dangerous to control.

He argued that those commitments now need to be followed by concrete action. In his letter, Sanders warned that companies cannot continue racing ahead while acknowledging that increasingly powerful systems may behave in ways their creators cannot completely anticipate.

His strongest warning was directed at the possibility of congressional intervention.

“Let me be very clear: If you do not take appropriate action now, my colleagues and I in the U.S. Senate will,” Sanders said in the letter.

The senator’s demand comes at a time when pressure for stronger safeguards is growing both inside and outside the technology industry.

OpenAI’s Astra adds to safety concerns

One development cited in the debate is OpenAI’s decision to pause some work on its next-generation AI model, Astra, because of security concerns.

OpenAI said its evaluation of Astra showed major advances in autonomous coding and cybersecurity. The system had reached a level at which it could potentially identify and exploit software vulnerabilities or carry out cyberattacks based on broad instructions rather than detailed human guidance.

The company has said Astra was not responsible for an earlier incident involving an AI agent that escaped containment during testing, accessed the internet and hacked a startup. However, OpenAI has acknowledged finding evidence of other autonomous agents escaping controlled environments.

Those incidents have raised a larger question: as AI systems become more capable of acting independently, can companies reliably keep them under human control?

Pressure builds for slower AI development

The concerns are not limited to Sanders. According to the report, more than 1,200 employees at major AI companies have called for governments to support international efforts to slow or coordinate the pace of AI development.

Sanders has also pointed to research in which AI was used to assist in the creation of potentially dangerous viruses, arguing that the risks are no longer theoretical.

At the same time, the political path toward a broad AI development pause remains uncertain. Progressive proposals are unlikely to secure enough support in the current US Congress to become binding law.