AI Firms Begin Saying 'No' to High-Risk Model Releases and Military Use

One of the more encouraging developments in artificial intelligence is that some of the people building it have started acting like it might be dangerous. Not in the Skynet sense or the HAL 9000 sense—or even the "oops, it deleted all my emails" sense—but AI may indeed pose risks to infrastructure, privacy, security, and the blurred line between public and private domains.

AI Companies Take a Cautious Approach

For years, Big Tech has prioritized rapid innovation over caution, sparking debates about AI’s downsides while delivering significant benefits. However, in a notable shift, some companies are now saying no to unrestricted releases of powerful models.

Anthropic’s Project Glasswing: A Case Study in Restraint

Anthropic has announced it will not broadly release Claude Mythos Preview, a frontier model that reportedly identified thousands of high-severity vulnerabilities, including flaws in every major operating system and web browser. Instead, access is restricted to a consortium of critical infrastructure partners, including:

• Amazon Web Services
• Apple
• Broadcom
• Cisco
• CrowdStrike
• Google
• JPMorganChase
• Linux Foundation
• Microsoft
• NVIDIA
• Palo Alto Networks

The goal of Project Glasswing is defensive: to identify and patch catastrophic flaws before malicious actors exploit them. While the move serves corporate interests—positioning these firms as responsible stewards rather than reckless innovators—it marks a departure from the industry’s long-standing mantra of "ship first, ask questions later."

Anthropic Draws a Line with the Pentagon

Anthropic’s caution extended to military applications. The company publicly stated it would allow only two narrow exceptions for military use of its models: mass domestic surveillance and fully autonomous weapons.

On surveillance, CEO Dario Amodei warned that AI could turn commercially available data into "a comprehensive picture of any person's life—automatically and at massive scale."

Regarding autonomous weapons, Amodei argued that today’s frontier systems are "not reliable enough" to "take humans out of the loop entirely and automate selecting and engaging targets".

While Anthropic remains willing to assist the government with conventional military applications, its refusal to waive restrictions on surveillance and autonomous weapons drew a fierce response from the Pentagon.

Pentagon Responds with Heavy-Handed Tactics

The Defense Department insisted it would contract only with AI companies willing to accept "any lawful use" and remove safeguards. When Anthropic refused, Defense Secretary Pete Hegseth designated the company a risk to national security, a label that