Officials in the Trump administration are working on an executive order that would push federal agencies to partner with artificial intelligence companies to defend networks against cyber attacks that leverage AI, according to a report from Bloomberg News citing people familiar with the matter. Those sources discussed the draft order on the condition of anonymity because the deliberations remain confidential.
The proposed directive would revise current cybersecurity information-sharing programs to explicitly include AI companies, the people said. The aim, as described by those familiar with the plan, is to address threats that arise from the emerging technology by improving how information about vulnerabilities and incidents is exchanged among stakeholders.
People briefed on the draft added that the order would not establish a requirement for government approval of advanced AI models. In other words, the changes contemplated would stop short of imposing a pre-approval regime on cutting-edge systems, while still bringing AI firms into established channels for cyber threat information.
According to the sources, the contemplated updates are intended to make it easier to identify and repair weaknesses across networks at the federal, state and local levels, as well as within critical U.S. infrastructure. At the same time, the administration intends to avoid creating new oversight requirements for AI models.
Summary
The administration is preparing an order to have U.S. agencies work with AI companies to defend against AI-enabled cyber attacks. The draft would amend existing information-sharing programs to include AI firms, help find and fix vulnerabilities across government and critical infrastructure networks, and would not require government sign-off for advanced AI models.
Key points
- The draft executive order would direct federal agencies to collaborate with AI companies on network defense and cybersecurity information-sharing.
- The changes would broaden existing programs to include AI firms and focus on identifying and remediating vulnerabilities across federal, state and local networks and critical U.S. infrastructure.
- The directive would stop short of mandating government approval for cutting-edge AI models, avoiding a pre-approval oversight regime.
Risks and uncertainties
- The specifics of how the information-sharing updates would be implemented remain unclear; sources described the draft but spoke on condition of anonymity.
- By not requiring government approval for advanced models, the directive leaves open questions about the scope of oversight applied to cutting-edge AI systems.
- While the order aims to cover federal, state and local networks and critical infrastructure, the effectiveness of the updates in practice is not detailed in the available reporting.