Executive Order 14110 on safe, secure and trustworthy AI signed

Executive Order 14110, Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, was signed on 30 October 2023. It launched a coordinated, government-wide federal effort to govern how AI is developed and used.

What it is

The order rested on eight guiding principles, the first of which was that AI must be safe and secure. Meeting that principle, it said, required robust, repeatable and standardized evaluations of AI systems before they are put to use.

It also wrote a working definition of AI red-teaming into federal policy: structured testing to find flaws and vulnerabilities in an AI system, often in a controlled environment and alongside the system's developers.

What it requires

Section 4.1 gave the Commerce Department, acting through NIST, 270 days to issue guidelines and best practices, including a generative AI companion to the AI Risk Management Framework and procedures for developers of dual-use foundation models to run red-team tests.

Section 4.2 invoked the Defense Production Act. It told Commerce to require companies developing potential dual-use foundation models to report on an ongoing basis, including on who owns the model weights and the physical and cybersecurity measures that protect them.

The order no longer applies. Executive Order 14148, signed on 20 January 2025, lists EO 14110 among the actions it revoked.

What security leaders should do

• Keep the guidance the order produced on your reading list; documents such as NIST's generative AI profile stand on their technical merits even after the order's revocation.
• Adopt the order's red-teaming definition in internal testing policy, because it is precise and vendor-neutral.
• Where contracts or policies still cite EO 14110, re-anchor them to the underlying standard or guidance so obligations do not point at a revoked instrument.
• Protect model weights as crown-jewel assets whether or not a reporting rule applies to you.