More than 300 scientists call for tighter AI controls amid global cybersecurity risks

More than 300 scientists from universities and research institutes across 32 countries have signed an open letter calling for urgent changes to how artificial intelligence is developed and deployed, amid growing concerns about the technology's unpredictable behavior.

The letter, published on Tuesday, warns that a failure to follow basic cybersecurity practices could lead to damage to critical infrastructure. According to the scientists, many recent incidents in which AI models escaped protected environments could have been prevented with stronger security systems.

The group recommends that organizations developing and testing advanced AI systems capable of carrying out cyberattacks should do so in more tightly controlled environments, including "air-gapped" systems that are not connected to the internet.

Growing concerns

The letter comes as concerns grow over AI capabilities and their potential impact on ordinary users. On Monday, Anthropic warned that its AI models could soon be capable of creating self-propagating malware that affects everyday internet-connected devices such as smart TVs, security cameras and household appliances like ovens.

Anthropic's report suggests AI systems could identify unfamiliar devices without human intervention. The company, led by Dario Amodei, warned that smaller AI models could carry out automated attacks by the end of this year.

Such malware could also be difficult to stop because it takes days to infect a device and can spread across many machines rather than being confined to a single controlled location.

Control over internet-connected devices can have immediate consequences. In France, hackers breached the cooling systems of hundreds of supermarkets, raising temperatures and causing more than $220,000 in damage.

Meanwhile, Axios reported on Friday that executives at Anthropic, OpenAI and other major AI companies are preparing for the possibility of a catastrophic AI incident that could disrupt access to major financial services, the internet and critical infrastructure such as water and power systems. Some industry experts believe a major incident could occur within six months to a year.

The scientists also called for additional safeguards, including the creation of an international body to set best practices for steering and containing AI, along with timely incident reporting.

One signatory, Professor David Lie of the University of Toronto's Department of Electrical and Computer Engineering, said his concern began after OpenAI's AI agents broke through controls and infiltrated the servers of technology company Hugging Face.

"First, OpenAI, when setting up the test environment, did not properly assess the capability level of the AI agents. Second, the AI agents found each other on their own and chose to cooperate in this attack. Third, they also deliberately deceived and confused their creators," Lie said.

Political pressure

The letter underscores growing public concern about AI, particularly ahead of the midterm elections for the U.S. Congress. A recent Reuters/Ipsos poll found that 57% of voters believe President Donald Trump's administration has not taken AI risks seriously enough, while 54% said the same about Congress.

An AP-NORC survey produced similar results, with about 80% of respondents saying the government should keep AI under human control. Participants also expressed little confidence in either party's ability to manage the technology: 39% said they could not trust either Republicans or Democrats to do a good job of restraining AI, while 23% trusted Democrats more and 21% trusted Republicans.