On June 2, 2026, President Trump signed an executive order titled “Advancing Innovation and Security in Advanced Artificial Intelligence.” This is the administration’s most significant step toward federal oversight of AI centered almost entirely on cybersecurity. See Execution. “Advancing Advanced Artificial Intelligence Innovation and Security” Order (June 2, 2026).
I. Important points
- Orders are voluntary and not mandatory. There will be no licensing or pre-approval requirements for AI development. See Execution. Order § 3(c) (“Nothing in this section shall be construed as authorizing the creation of mandatory government licensing, prior approval, or permitting requirements for the development, publication, release, or distribution of new AI models, including Frontier Models.”)
- This order establishes three important institutional frameworks: (1) a Treasury-led AI Cybersecurity Clearinghouse, (2) a confidential benchmarking process to designate “covered Frontier Models,” and (3) a voluntary 30-day pre-release review period for Frontier AI models. See Execution. §§ 2(d), 3(a) through (b).
- The directive is silent on privacy, data protection, and data subject rights.
- For organizations outside of the AI development space, a more pressing question is whether and how these changes will impact data privacy posture, vendors, and regulatory risks.
II. Key definition
This mandate creates a new label for the most powerful AI systems called “covered frontier models.” See Execution. Order §3(a). The EO itself does not substantively define this term. Rather, it directs authorities to develop a confidential benchmarking process to “assess the advanced cyber capabilities of AI models and determine the threshold at which an AI model should be designated as a ‘covered frontier model.'” However, the Covered Frontier model appears to be a highly capable AI system that can pose significant cybersecurity risks, including finding and exploiting weaknesses in software on its own. NSA conducts a confidential process to determine which models are eligible, in consultation with the Director of the National Cyber Bureau, the Assistant to the President for Science and Technology (APST), the Director of CISA, and other representatives of the Department of the Army as appropriate. Executive Order §3(a). If a model receives that designation, developers can voluntarily submit it to government review before release.
III. What an order actually does
Almost all directives have a 30 or 60 day deadline, so orders are placed quickly. It does four main things:
- He directed CISA to strengthen the federal government’s cyber defenses and share AI-powered security tools with state and local governments and critical infrastructure operators. See Execution. Order §§ 2(c)(i)–(iii);
- Tasks the Treasury Department with creating an “AI Cybersecurity Clearinghouse” that will work with AI companies and infrastructure operators to coordinate the discovery and patching of software vulnerabilities. See Execution. Order §2(d);
- We are providing a voluntary 30-day period for AI developers to submit their most powerful models to government cybersecurity reviews before release. See Execution. Order § 3(b)(ii);
- Directed the Attorney General to prioritize criminal enforcement against those who use AI to infiltrate systems or steal data. See Execution. Order §4.
IV. Why this is important for data privacy even though it doesn’t say “privacy””
The executive order does not address algorithmic bias, the impact of AI on employment, transparency, or data subject rights. This order focuses solely on national security and cybersecurity. The Trump administration appears to have changed course on AI after Anthropic’s Claude Mythos Preview model, announced in April, was demonstrated to be able to autonomously identify and exploit hidden vulnerabilities in widely used software. [Note: The causal link between the Mythos announcement and this EO is an inference based on timing; no official source has confirmed this connection.]
These “covered frontier models,” systems that can find security holes in software, can also reach the personal data the software is supposed to protect. Every vulnerability is a potential gateway to private information. While the order talks about “hardening systems” and “patching vulnerabilities,” it also talks about protecting personal data within those systems.
V. What hasn’t changed? Well, at least not yet.
At this time, there are no new privacy rights, data protection rules, or consent requirements for AI. Issues regarding training data, opt-out rights, and disclosure obligations are still governed by the existing patchwork of state privacy laws, HIPAA, GLBA, and FTC enforcement. The order is also completely voluntary, meaning open source and open weight models that researchers have shown can reproduce frontier-level hacking capabilities will not be captured and may flow into the supply chain without government review. And while clearinghouses can help discover vulnerabilities faster, getting organizations to actually patch them is always the difficult part, especially for data-rich organizations that don’t have well-resourced security teams. Finally, the directive does not address AI-specific privacy threats such as prompted injection, model inversion, and data poisoning, attacks in which personal data is directly exposed through security flaws.
VI. What to do now
Although the order is voluntary and does not directly regulate privacy, it does signal where the federal government’s attention is heading. Organizations, especially those that handle large amounts of personal data, should take the following steps now:
- Audit your AI vendor contracts. Check to see if any vendors have developed or introduced models that can be designated as “covered frontier models.” Update vendor contracts to require notification if a model receives that designation and specify whether the vendor will participate in the government’s voluntary review process.
- Pressure test your patch management program. A new clearinghouse means faster discovery of vulnerabilities and more patches delivered faster. If your organization is already struggling with patching frequency, now is the time to close the gap before clearinghouses start publishing findings that impact your systems.
- Map exposures to open source and open weight AI models. The executive order’s voluntary framework does not cover open source or open weight models, even if they reproduce frontier-level functionality. Review your software supply chain to identify where these models reside and assess the associated risks, especially if the models interact with systems that store or process personal data.
- Evaluate defenses against AI-specific privacy attacks. The executive order does not mention prompted injection, model inversion, or data poisoning. Don’t wait for guidance from the federal government. If you use AI systems that process personal data, evaluate whether your current security controls address these threat vectors and whether your incident response plans account for them.
- Monitor 30-day and 60-day deadlines. The executive order’s implementation schedule is quite aggressive. Key deliverables, including the clearinghouse framework and confidential benchmark standards, are expected to be submitted within one to two months. These outcomes will shape what the “covered frontier model” means in practice and may impact procurement decisions, vendor relationships, and compliance plans.
