OpenAI, Anthropic Limit New AI Models During Trump Cybersecurity Review/ Newslooks/ WASHINGTON/ J. Mansour/ Morning Edition/ OpenAI and Anthropic have limited access to their latest artificial intelligence models following cybersecurity reviews requested by the Trump administration. The temporary restrictions reflect growing government scrutiny over advanced AI systems capable of identifying software vulnerabilities and other national security risks.
OpenAI AI Cybersecurity Review Quick Looks
- OpenAI limits GPT-5.6 Sol to government-approved customers.
- Anthropic receives approval for limited deployment of Mythos 5.
- Trump administration is reviewing advanced AI models for cybersecurity risks.
- Officials fear powerful AI could be exploited for cyberattacks.
- Critics warn government intervention may slow U.S. AI innovation.
- AI oversight comes as both companies explore future public offerings.
Deep Look
Government Review Delays Public AI Rollouts
OpenAI and Anthropic have become the first major artificial intelligence companies to restrict access to newly developed AI models following cybersecurity reviews requested by the Trump administration.
OpenAI announced Friday that its newest flagship model, GPT-5.6 Sol, will initially be available only to a limited group of customers approved by the federal government. Anthropic later confirmed that regulators authorized a limited release of its advanced cybersecurity-focused model, Mythos 5, after previously restricting deployment.
Both companies described the move as temporary while federal officials complete security evaluations of the rapidly advancing technology.
OpenAI Limits GPT-5.6 Sol Access
OpenAI said GPT-5.6 Sol represents a significant leap in AI capabilities, particularly in identifying and fixing software vulnerabilities.
“We don’t believe this kind of government access process should become the long-term default,” OpenAI said in a statement. The company said it viewed the testing period as a temporary step on the “path to broader availability in the coming weeks.”
Although the company believes the model performs better at strengthening cybersecurity than conducting offensive attacks, executives acknowledged uncertainty surrounding its broader capabilities.
As a result, OpenAI agreed to release the model gradually under enhanced safeguards.
“That uncertainty, along with the model’s broader step change in capabilities, is why we are pairing the model’s increased capabilities with stronger safeguards and a phased release,” the company said Friday.
The company emphasized that government approval should not become the long-term standard for AI deployment and expects wider public availability in the coming weeks.
So far, approximately 20 organizations have received authorization to access GPT-5.6 Sol, though OpenAI has not identified those customers.
Anthropic Receives Partial Approval
Anthropic also announced progress after federal restrictions forced it to suspend two recently introduced AI models earlier this month.
The Trump administration approved the limited return of Mythos 5, allowing deployment to selected cybersecurity professionals and critical infrastructure operators.
“I just want to say that pretty much nobody in the cybersecurity industry believes that there’s any factual basis for this action,” Stanford University cybersecurity expert Alex Stamos said on a call with reporters earlier this week.
However, Anthropic’s companion model, Fable 5, remains unavailable to the general public despite company efforts to address government concerns.
Federal officials cited worries that Mythos possessed advanced cyber capabilities capable of identifying software vulnerabilities that malicious hackers could exploit against critical infrastructure.
Trump Administration Expands AI Oversight
The restrictions stem from President Donald Trump’s executive order signed earlier this month establishing a framework for reviewing the national security risks of advanced artificial intelligence systems before widespread public release.
The initiative allows the federal government to evaluate frontier AI models for up to 30 days before deployment.
While participation remains technically voluntary, AI companies have largely cooperated with the review process as federal officials seek safeguards against emerging cybersecurity threats.
The White House said Friday it continues working closely with leading AI developers to balance innovation with national security.
Cybersecurity Risks Drive Government Concerns
Federal scrutiny intensified after Anthropic disclosed earlier this year that its Mythos model had become exceptionally skilled at identifying software weaknesses.
Administration officials argued that such capabilities could be weaponized if powerful AI systems fell into the wrong hands.
David Sacks, who serves on President Trump’s technology advisory council, said Anthropic’s public warnings significantly raised concerns in Washington about AI becoming a potential cyber weapon.
OpenAI similarly acknowledged that combining GPT-5.6 Sol with other tools could create unforeseen security risks despite internal testing indicating the model remains below the company’s highest risk thresholds.
Critics Warn of Political Gatekeeping
Some lawmakers and cybersecurity experts criticized the administration’s approach, arguing that government approval of AI customers creates uncertainty and could undermine American technological leadership.
Rep. Lori Trahan, a Democrat from Massachusetts, questioned why political appointees should determine which organizations receive access to advanced AI systems without clear legal standards or oversight.
Stanford cybersecurity expert Alex Stamos also challenged the government’s actions, arguing there is little evidence that Anthropic’s restricted models present greater risks than AI products already available from competitors, including Chinese developers.
Critics warn that slowing domestic AI innovation could weaken the United States in its technological competition with China.
AI Companies Continue Working With Washington
Despite concerns over increased oversight, OpenAI and Anthropic continue negotiating with federal officials regarding future AI deployments.
OpenAI CEO Sam Altman recently met Commerce Secretary Howard Lutnick to discuss the release of GPT-5.6 Sol and broader AI policy issues.
Anthropic has maintained a more contentious relationship with the administration after raising repeated concerns about AI safety and military applications, leading to legal disputes over government restrictions.
Still, company officials expressed optimism Friday that additional access to Mythos and eventually Fable could be approved following continued cooperation with regulators.
Oversight Adds Complexity to AI Industry Growth
The new federal review process arrives as OpenAI and Anthropic continue exploring potential initial public offerings amid soaring investor interest in artificial intelligence.
Government oversight could become a significant factor in future valuations and product launch strategies as regulators seek greater involvement in managing advanced AI technologies.
President Trump has also floated the idea of the U.S. government owning stakes in major AI companies, suggesting Americans could eventually share directly in the industry’s future growth.
You must Register or Login to post a comment.