The Short Version
The Trump administration just signed an Executive Order titled "Promoting Advanced Artificial Intelligence Innovation and Security." The headline: The United States leads the world in AI because of the enormous talent and innovation of the AI industry and because the administration refuses to stifle that innovation with overly burdensome regulation. Translation? Washington is declaring itself pro-AI, anti-red-tape, and very much aware that China exists.
But here is the thing: Buried beneath the deregulatory fanfare are several mechanisms that could meaningfully reshape how AI companies go to market and how financial services firms govern their AI programs. Some of these are genuinely useful. Others are traps dressed up as opportunities. Read on.
What the Order Actually Does
“The Philosophy”
The administration frames its prior actions as having "unleashed tremendous technological growth and economic investment in AI by slashing the bureaucratic constraints that the prior administration placed on America's AI developers and researchers," while acknowledging that advanced AI capabilities also introduce new national security considerations that require coordinated action across executive departments and agencies. In other words: Full speed ahead, but maybe put a seatbelt on.
The stated policy is to promote AI innovation and security by working collaboratively with the private sector to modernize government and private sector information systems and harden them against external threats; to protect American ingenuity and intellectual property from exploitation and theft by adversaries; and to cultivate America's advanced AI-enabled capabilities. So: innovation, IP protection, and cybersecurity. Got it.
“The Mechanics”
- Cybersecurity Directives (30 days): The Secretary of Homeland Security through CISA must release Binding Operational Directives to expedite and prioritize the cyber defense of civilian federal government information systems, establish or expand federal programs and cybersecurity services that enhance AI-enabled defensive tools, and facilitate access to cybersecurity tools and services, including covered frontier models, for agencies, state and local authorities, and operators of critical infrastructure such as rural hospitals, community banks, and local utilities. Community banks get a namecheck. More on why that matters below.
- The Treasury AI Clearinghouse (30 days): The Secretary of the Treasury, in consultation with the National Cyber Director, NSA, and CISA, must form an AI cybersecurity clearinghouse, in voluntary collaboration with the AI industry and operators of critical infrastructure, that coordinates and deconflicts scanning for software vulnerabilities, discovers and validates such vulnerabilities, and coordinates and prioritizes remediation and distribution of vulnerability patches.
- Frontier Model Framework (60 days): The relevant agencies must develop and maintain a classified benchmarking process to assess the advanced cyber capabilities of AI models and determine the threshold at which an AI model should be designated a "covered frontier model," with such determinations made by the Director of NSA, in consultation with the National Cyber Director, the APST, the Director of CISA, and other representatives of the Department of War. The voluntary framework that follows would allow developers to engage the federal government to determine whether their models qualify as covered frontier models, provide the government with access to those models for up to 30 days before planned release to trusted partners (subject to confidentiality and IP protections), and collaborate with the government to select trusted partners who receive early access.
- The One Protection That Actually Matters: Nothing in this section shall be construed to authorize the creation of a mandatory governmental licensing, preclearance, or permitting requirement for the development, publication, release, or distribution of new AI models, including frontier models. No federal approval gate before you can ship your model. Write that down.
- Criminal Enforcement (Immediate): The Attorney General is directed to prioritize the enforcement of 18 U.S.C. 1028, 18 U.S.C. 1030, 18 U.S.C. 1343, and all other applicable federal criminal laws against anyone who utilizes AI to illegally access or damage a computer without authorization, or who employs AI agents to unlawfully access data or information that is subsequently used for a criminal or unlawful purpose.
What the Order Gets Right
1. No Pre-Clearance. Full Stop.
Let's say it again because it deserves repeating: Nothing in this order shall be construed to authorize the creation of a mandatory governmental licensing, preclearance, or permitting requirement for the development, publication, release, or distribution of new AI models, including frontier models. In a world where the EU AI Act has introduced conformity assessments and regulatory sandboxes, the explicit rejection of a federal approval gate is a material competitive advantage for U.S. AI companies. This is real, and it matters.
2. Community Banks Finally Get Federal Backup
Community banks are explicitly identified alongside rural hospitals and local utilities as operators of critical infrastructure that should receive facilitated access to cybersecurity tools and services, including covered frontier models. This is not window dressing. It formally positions the community banking sector as a federal cybersecurity priority which means government-backed AI tools, vulnerability patching support, and a seat at the clearinghouse table. For years, smaller financial institutions have been outgunned on cybersecurity relative to their too-big-to-fail peers. This order nudges the dial.
3. Treasury at the Helm of the Clearinghouse Is Smart Architecture
Putting the Treasury in charge of the AI cybersecurity clearinghouse is a deliberate choice that embeds financial services into the framework's DNA. The clearinghouse is to be formed in voluntary collaboration with the AI industry and operators of critical infrastructure, coordinating and deconflicting vulnerability scanning, validating vulnerabilities, and prioritizing remediation and patch distribution. Treasury's proximity to bank regulators means financial institutions are not an afterthought in how this body develops, but instead, they are the primary constituency. Early participants will shape the norms before the norms shape them.
4. IP Protection Gets Federal Backing
The protection of American ingenuity and intellectual property from exploitation and theft by adversaries is explicitly enshrined as a core policy objective of the United States. For AI companies with crown jewels as model weights, training data, and proprietary architectures, a federal administration with a strong IP enforcement posture is a meaningful tailwind.
5. The Administration Acknowledges AI Is a Security Issue, Not Just an Economic One
The order commits that the administration will continue to work closely with the AI industry to ensure that the best and most secure technology is deployed rapidly to confront any and all threats to the country, leading an America-first cybersecurity effort that enhances both national security and global AI dominance. There is genuine value in the federal government treating AI security as a shared public-private challenge rather than a compliance problem to be delegated entirely to the industry. That framing, at least, is correct.
The Red Flags (Yes, There Are Several)
Red Flag 1: The Frontier Model Designation Is a Black Box
Here is the one that should keep AI company counsel awake at night. The benchmarking process used to determine whether a model qualifies as a "covered frontier model" is classified, with assessments shared with AI developers and researchers only "as appropriate," and determinations made by the director of the NSA in consultation with the National Cyber director, the APST, and other representatives of the Department of War. Translation: The government will decide if your model crosses a threshold you cannot see, using criteria you cannot review, and tell you about it when they feel like it. There is no published standard, no notice-and-comment process, and no procedural challenge mechanism established by this order. For companies with advanced agentic or code-generation models in development, this is not a hypothetical risk; it is a live one.
Red Flag 2: "Voluntary" Is Doing a Lot of Heavy Lifting
The order is emphatic that nothing in this section shall be construed to authorize the creation of a mandatory governmental licensing, preclearance, or permitting requirement for the development, publication, release, or distribution of new AI models, including frontier models. Good. But read the next part carefully.
Participating developers collaborate with the federal government to select trusted partners that will have early access to covered frontier models to promote secure innovation and strengthen the cybersecurity of critical infrastructure. So the companies that opt in get to decide who else gets early access. The companies that opt out? They're excluded from that trusted partner network. For AI vendors with primary customers being financial institutions (institutions which will themselves be seeking covered frontier model access), non-participation is commercially costly even if it is legally optional. Call it what you want, but "voluntary" with real market consequences is a peculiar kind of voluntary.
Red Flag 3: The IP Protections Don't Exist Yet
The order promises that developers who provide pre-release model access will receive appropriate confidentiality, cybersecurity, insider-risk, and intellectual-property protection, use, and nondisclosure requirements. “Appropriate” is doing enormous work in that sentence. Those protections are to be designed as part of the framework, which does not yet exist as a legally binding instrument. Until those terms are finalized, codified, and reviewed by counsel, providing the government with access to your model weights is betting your most valuable IP on a handshake deal with the federal government. And to be clear: This order does not, and does not intend to, create any right or benefit, substantive or procedural, enforceable at law or in equity by any party against the United States, its departments, agencies, or entities, its officers, employees, or agents, or any other person. So, if the protections prove inadequate, there is no recourse. Like, none.
Red Flag 4: The Clearinghouse Could Be a Trojan Horse
The clearinghouse will coordinate scanning for software vulnerabilities, discover and validate those vulnerabilities, and coordinate and prioritize remediation and distribution of vulnerability patches. Useful, in theory. But participating in vulnerability scanning requires disclosing architectural details and system information about your AI systems to a government-convened body. The order contains no provision establishing what the clearinghouse does with that information once it has it. No confidentiality framework, no data handling rules, no limitations on secondary use. Companies that rush to participate before those terms are established are handing over sensitive system information on terms they have not reviewed.
Red Flag 5: The CFAA Provision Is Broader Than the Obvious Reading
The attorney general is directed to prioritize enforcement against anyone employing AI agents to unlawfully access data or information that is subsequently used for a criminal or unlawful purpose. The "subsequently used for a criminal or unlawful purpose" tail deserves a closer look. It potentially attaches criminal liability to AI-enabled data access that appeared authorized at the time, if someone downstream uses that data unlawfully. Financial institutions deploying AI agents across third-party data sources. AI vendors whose tools end up in a bad actor's hands face real exposure here. The Computer Fraud and Abuse Act has always been drafted broadly. Coupling it with AI and prioritized enforcement is a combination that demands careful legal review of every AI agent's data access scope.
Red Flag 6: This Order Does Not Touch Your Real Regulators
This is the one financial services firms are most likely to get wrong. Nothing in this order shall be construed to impair or otherwise affect the authority granted by law to an executive department or agency, or the head thereof. The SEC, OCC, FDIC, CFPB, and FINRA are completely unaffected. This order's deregulatory rhetoric means precisely nothing to the examiners showing up to review your model risk management program. Companies that interpret "America is pro-AI" as a signal to dial back compliance investment in AI governance are setting themselves up for a rude awakening the next time a bank examiner asks about their explainability documentation.
Red Flag 7: None of This May Actually Happen on Schedule (It Probably Won’t)
This order shall be implemented consistent with applicable law and subject to the availability of appropriations. The clearinghouse, the frontier model framework, and the expanded cybersecurity services are all contingent on Congress providing the money. If appropriations are not forthcoming, or are delayed, those 30- and 60-day deadlines are aspirational at best. And we all know what Congress’ most recent track record looks like.
Next Steps: What To Do Before Monday Morning
If You Are a Financial Services Company
- Get your AI inventory in order. CISA's Binding Operational Directives, due within 30 days, will set cybersecurity standards that historically migrate from federal systems to regulated industries faster than most compliance teams are ready for. Know what AI you are running before Washington tells you what it expects.
- Show up to the clearinghouse (if you have the appetite), but read the terms first. The Treasury-led clearinghouse is a genuine opportunity to shape vulnerability disclosure norms for the financial sector. It is also a forum where you will be asked to share sensitive system information under a framework with confidentiality rules that do not yet exist. Get counsel to review participation terms before you submit anything.
- Do not mistake this order for regulatory relief. Your real regulators (the ones with exam authority) are completely untouched by this order. AI governance investment, model risk management, third-party oversight, or explainability documentation: None of that goes away because the White House likes AI. Keep spending.
- Audit your AI agents' data access scope now. The CFAA enforcement priority creates a new lens on any AI agent that touches third-party data. If the authorization scope is ambiguous, resolve it. Document it. Do not wait for enforcement to clarify the question for you.
- Review every AI vendor agreement. In light of the criminal enforcement provisions and the downstream liability risks they create, vendor agreements need a fresh look for: limitations on authorized use, responsibility allocation for misuse, indemnification, and security obligations. If your vendor's product ends up facilitating unauthorized access, you want the contractual framework to have contemplated that possibility.
If You Are an AI Company Serving Financial Services
- Figure out whether you could be a covered frontier model before the government decides for you. The voluntary engagement mechanism exists precisely for this purpose. Developers can engage the federal government to determine whether their models under development meet the designation of "covered frontier model." Use it proactively. Being blindsided by a designation in the middle of a product launch is considerably worse than engaging early on your own terms.
- Do not provide pre-release model access until the IP protections are in writing and reviewed. The promise of appropriate confidentiality, cybersecurity, insider-risk, and intellectual-property protection, use, and nondisclosure requirements is exactly that: a promise. Until those terms exist as a legally enforceable instrument, your model weights are not protected by anything in this order. The order explicitly creates no right or benefit enforceable at law or in equity against the United States or any of its officers, employees, or agents. That is not a legal foundation on which to expose your core IP.
- Engage with the clearinghouse strategically, but also not yet. The clearinghouse is a real opportunity to demonstrate security credibility to financial institution clients who increasingly treat vendor cybersecurity posture as a procurement criterion. But participate only once the confidentiality framework governing submitted information is clearly defined and reviewed. Being first through the door is only valuable if the door doesn't lead somewhere you did not intend to go.
- Update your customer agreements for the CFAA moment. If your product is used by a client or compromised by a third party to unlawfully access data, the enforcement priority directed at the attorney general creates potential reputational and legal exposure that your current agreements may not adequately address. Tighten limitations on authorized use, harden indemnification provisions, and make sure your incident response obligations are clear.
- Watch CISA's binding operational directives like a hawk. Those directives will establish or expand federal programs and cybersecurity services that enhance AI-enabled defensive tools and facilitate access to cybersecurity tools for agencies, state, and local authorities, as well as critical infrastructure operators. They will become the de facto standard that your financial institution clients benchmark against when evaluating vendor security. Know what is coming before your clients start asking whether you meet it.
The Bottom Line
The administration has committed to working closely with industry to ensure that the best and most secure technology is deployed rapidly to confront any and all threats to the country. That is a worthy ambition. Whether or not the mechanisms this order puts in motion (a classified designation process, an unfunded clearinghouse, a voluntary framework with undefined IP protections, and an expansive criminal enforcement mandate) will actually be delivered on will depend almost entirely on the implementation details yet to be written.
The policy of the United States is to promote AI innovation and security by working collaboratively with the private sector. "Collaboratively" is the operative word. The companies that engage early, help define the frameworks, and build relationships with the relevant agencies will be positioned to shape what this order becomes in practice. The companies that wait for the frameworks to arrive fully formed or that misread the deregulatory tone as an invitation to reduce compliance vigilance will find themselves reacting to standards they had every opportunity to influence. The frontier is being defined. If you’re hungry, grab a seat at the table.