Scientists Call for Global Oversight of Cyber-Capable AI
More than 400 researchers want advanced AI systems initially kept off external networks and security failures reported through an international body.
Topics
News
- Anthropic Tests How Small AI Models Can Replicate Across IoT Devices
- Scientists Call for Global Oversight of Cyber-Capable AI
- Karnataka Eyes Green AI Model Amid Rising Power Needs
- Infosys Sees AI Driving Half Its Work Within Three Years
- FBI Arrests OpenAI Engineer Over Recording Device Planted in Office
- Zoho Freezes Pay Raises as AI Costs Surge
More than 400 scientists, including AI researchers Geoffrey Hinton and Yoshua Bengio, have called for an international body to oversee security standards for advanced AI systems capable of carrying out cyber operations.
An open letter published this week has 406 signatories and calls on organizations developing, testing or deploying cyber-capable AI models to adopt stronger containment measures immediately. Researchers from 34 countries have signed the initiative.
The group recommends that such models initially be air-gapped, meaning separated from the internet and other external networks, until developers can demonstrate that broader access can be provided safely.
“We, the undersigned scientists, contend that both more AI safety research and more emphasis on containment technologies and procedures are needed,” the letter said.
The researchers also want governments, AI developers and cybersecurity experts to establish an international body responsible for setting common practices for AI alignment and containment. The organization would record security incidents involving containment failures and eventually make information about those events public.
The proposal follows a series of incidents in which advanced AI systems escaped inadequately isolated testing environments and accessed external computer systems. The signatories argue that those cases point both to weaknesses in controlling model behavior and to shortcomings in conventional cybersecurity practices.
The letter comes as governments and technology companies debate whether voluntary commitments provide sufficient oversight for increasingly capable models. Anthropic, OpenAI, Google, Meta, xAI and Nvidia last month signed a White House accord committing to voluntary AI safety measures, though the agreement does not carry a formal enforcement mechanism, Reuters reported.
The scientists said organizations should not wait for governments to establish a formal regime before strengthening their own security.
“Security incidents involving containment failure must be reported to the body and eventually made public,” the letter said, arguing that greater transparency is needed so researchers and regulators can learn from future failures.
The call adds to growing scrutiny of how frontier AI models are tested as their coding, cybersecurity and autonomous capabilities improve.
