President Trump has signed an executive order requesting AI companies to show models to the federal government, allowing the government to assess their capabilities prior to a full release.

Below, security leaders share their thoughts on the implications of his policy. 

Security Leaders Weigh In

Ram Varadarajan, CEO at Acalvio:

As attackers use AI to automate attacks, they move faster in gaining access and spreading inside the network; defenses built for human response times fail silently. CISOs investing in AI-native security aren’t chasing efficiency. They’re closing a fundamental speed gap between attack and defense.

The formalization of government pre-release reviews is marking the end of AI’s ‘Wild West’ era. Geopolitical alignment and national security clearances are going to become as critical to a frontier lab’s valuation as its raw compute.  

It’s a transition that’s going to transform frontier AI from a pure-play tech bet into a regulated strategic industry. We’re just seeing the first steps, and I have to say that it was bound to happen given the stakes at play.

Marcus Fowler, CEO of Darktrace Federal:

Darktrace Federal welcomes the Administration’s continued focus on the cybersecurity implications of advanced AI and recognition of the critical role AI-powered cybersecurity will play in defending federal networks, critical infrastructure, and State and local authorities. As cyber threats continue to increase in speed, scale, and sophistication, defensive AI is becoming an essential capability for government agencies seeking to strengthen mission resilience and maintain a security advantage over AI-powered attackers.

The next challenge is ensuring AI systems are deployed securely once they move into real operational environments. As AI becomes embedded across applications, cloud environments, autonomous agents, operational technology, and critical infrastructure workflows, organizations will need clearer visibility into how those systems behave, what data and resources they can access, and when activity moves outside expected parameters.

The security conversation must extend beyond model development and testing to focus on the operational realities of AI deployment. NIST’s AI Agent Standards Initiative and forthcoming guidance from CISA and other federal stakeholders will be important in helping organizations establish practical frameworks for securing AI in production environments, including how AI systems and agents are identified, authorized, monitored, and governed throughout their lifecycle.

Securing AI requires securing the wider ecosystem around it: the infrastructure, identities, data, networks, applications, and operational environments where AI systems increasingly act at machine speed. The goal should be to support AI adoption while giving organizations the visibility, control, and confidence needed to manage new risks as they emerge.

Dave Gerry, CEO at Bugcrowd:

Across every industry, from criminal gangs to nation-state actors, attackers are utilizing AI to accelerate their pace and frequency of attacks, increasingly causing defenders to be outmatched like never before. Whether through internal security teams or outsourcing part of their security operations to managed services firms, security teams must rapidly ramp up their usage of AI in response to the increased threat environment. 

Anytime that an administration is publicly prioritizing cybersecurity at a very strategic level is a positive sign for the industry and for broader national security implications. It’s a meaningful first step. 

Today, the biggest gap in the U.S. government’s approach to disrupting global cybercrime operations is speed.  Adversaries move faster than the government. This is just the reality of the environment we’re in and AI has only amplified this fact. This forces the government to be in a constant state of catch up. 

The majority of federal cybersecurity policy is based on compliance frameworks, post-breach policy and incident response instead of proactive vulnerability discovery to avoid the issue before it happens. While we’ve seen things like bug bounty and vulnerability disclosure programs be successful all across the federal government, they’re still not standard practice or required for every agency or critical infrastructure operators. 

State and local governments are falling behind in terms of capability, capacity and funding. The federal programs get the attention and funding, but the cybercriminal groups are disproportionately targeting smaller, less sophisticated organizations. The same is happening in the private sector across large versus small healthcare systems, large versus small utilities, etc. 

The biggest gap isn’t in strategy, it’s in the speed of operating. Adversaries today are operating at machine speed and the government is operating at bureaucracy speed. Proactive security must become the default to help offset this velocity gap.

Robert Costello, Chief Digital and Information Officer at Merlin Group:

The pace of AI advancement is eclipsing anything we saw in previous technology revolutions, so it’s encouraging to see American AI companies working collaboratively with the Trump administration to balance cyber safety with rapid innovation that helps maintain our technology superiority. The current review period is a tremendously positive step, giving the federal government a meaningful window to assess upcoming releases and work with cyber industry counterparts on concerns before they become problems. I look forward to seeing how this plays out over the coming months.

Diana Kelley, Chief Information Security Officer at Noma Security:

Voluntary security programs can work, but only when they create real accountability. We’ve seen this in cyber before. Coordinated vulnerability disclosure began largely as voluntary cooperation between researchers and vendors, but it became more effective when organizations added clear intake channels, response timelines, safe harbor language and public accountability. Post-incident review models such as the Cyber Safety Review Board are also useful: they don’t regulate directly, but they can still create pressure, shared lessons and concrete recommendations. Industry frameworks like the NIST Cybersecurity Framework and the Secure Software Development Framework are also voluntary in many contexts, but they gain teeth when procurement, audits, insurers, customers and regulators start expecting them.

For frontier AI, a 90-day government review could be useful as one checkpoint, but evaluating model safety is complex and ongoing. The risks evolve after release, especially when models are connected to agents, code execution, enterprise data, identity systems or critical infrastructure workflows. Review needs to account for how the model is deployed, what it can access, how much autonomy it has, and what guardrails are actually enforced in production.

The big question is whether this helps establish a durable safety assessment process, one that includes independent testing, clear risk thresholds, disclosure obligations, post-release monitoring, incident reporting and meaningful consequences when unacceptable risks are found. Without that structure, a voluntary process could look reassuring without materially reducing risk.

Rajeev Gupta, Co-Founder & CPO at Cowbell:

The bigger issue is that the government simply isn’t equipped to meaningfully oversee frontier AI models on its own. Even with a 90-day review window, it’s unclear which agency would have the technical expertise and staffing needed to properly evaluate these systems at the pace AI is advancing.

A more effective model would be a public-private consortium where leading AI labs contribute funding, talent, and technical resources, while the government provides regulatory authority and enforcement. There’s precedent for this approach: after the Three Mile Island incident, the nuclear industry created the Institute of Nuclear Power Operations (INPO), which ultimately became more rigorous in enforcing safety standards than regulators alone.

AI may require a similar framework. Supporting an independent body that helps ensure accountability should be viewed as a core cost of operating at frontier scale, and not just as a regulatory burden.

Collin Hogue-Spears, Senior Director of Solution Management at Black Duck:

The voluntary 90-day review matters because it is the cleanest and fastest legal path for pre-release AI model scrutiny while the administration’s broader legislative framework sits with Congress.

Voluntary is not the policy floor. It is the legal ceiling on executive AI review without Congress. China required generative-AI service filings in 2023 through their Cyberspace Administration of China rules. The European Union made general-purpose AI documentation and cooperation obligations applicable in August 2025 under the AI Act. The U.S. is building a voluntary review lane because existing national-security statutes offer no obvious basis for compelled model submission.

The administration already sent Congress a March 2026 AI legislative framework that calls for federal preemption of burdensome state AI laws. That framework has not become binding law. The Center for AI Standards and Innovation (CAISI), a division of NIST, already had voluntary testing agreements with Google, Microsoft, and xAI before this order. Mythos accelerated the administration’s return to pre-release model scrutiny, but the executive order expands the national-security audience, not the legal authority. It does not turn voluntary testing into a binding regime, and it does not create a national AI standard, and it does not displace the state-by-state rules already forming in Colorado, California, New York, Texas, and Virginia.

The unresolved policy question is whether Congress links pre-release AI review to procurement eligibility, export approvals, or both. Until then, the U.S. has a voluntary review lane while China and the European Union are shaping global AI governance conversations and defining the standards companies must build around. The U.S. framework has testing capacity, but it is still governing through voluntary review, executive orders, and a stalled legislative framework. Until Congress acts, voluntary review will not become market access, and federal policy will not preempt the state AI patchwork, which will governance increasingly harder for every U.S. AI company shipping across state lines.

Randolph Barr, Chief Information Security Officer at Cequence Security:

About two-thirds of current AI-related incidents still originate from traditional weaknesses, however, the remaining third are uniquely “AI-native.” These include model and data poisoning, prompt injection, and autonomous agents that can chain together API calls and act with minimal human oversight. These emerging risks reflect the reality that AI systems are dynamic, self-learning, and interconnected in ways traditional applications never were. When paired with the rapid speed of development, the outcome is a growing attack surface that grows faster than most security programs can respond. 

We are approaching a future where the use of AI agents will outpace the readiness of security measures. We have seen a number of advisories over the past year which help highlight the gaps and hopefully drive the industry toward more secure, transparent designs before these tools become deeply embedded in enterprise ecosystems.

John Gallagher, Vice President at Viakoo:

We are still in the early stages of defining what constitutes “safe use.” To be clear, advanced AI itself is a massive supply chain risk. If these “frontier” models are eventually integrated into operational technology (OT) or physical security systems (like smart cameras or building controllers), the integrity of the model itself becomes a critical OT security concern. Ensuring that an AI agent managing a physical network hasn’t been “poisoned” or tricked into disabling security protocols is the next step in establishing the risk of using AI-enhanced OT systems.

Anthropic’s Claude Mythos is but one “Frontier AI,” and Project Glasswing is but one (narrow IT-focused) effort in reducing the risk of general availability. By working with all advanced AI models pre-production the hope is much more broad efforts can be undertaken, such as testing against OT systems and establishing use-case specific guardrails.