Anthropic Walked the Walk

The refusal of Anthropic to waiver under political pressure and compromise its AI safeguards can lead to improved employee morale, increased productivity, attracting and retaining talent, gaining news business and engendering trust. The post Anthropic Walked …

The recent controversy over Anthropic’s decision not to cede to the Defense Department’s demands to remove its safety guards on Claude, its AI software, is an interesting look at a company whose purpose seems to go beyond just making money. The Pentagon—who Anthropic held a contract with for approximately $200 million—demanded policy safeguards be removed. CEO Dario Amodei is afraid AI could be used by the military to create fully autonomous weapons that could strike a target without any human intervention.

He also fears the AI platform could potentially be used for mass domestic surveillance. Beyond the moral questions, it should be noted there are currently few laws in place that determine what the government can and can’t do with artificial intelligence, as reported by Axios. Amodei’s public statement on Feb. 26 noted it’s the Defense Department that makes military decisions, not private companies.

But he drew a red line, saying “…we cannot in good conscience accede to their request." Beyond the projected financial costs (although Anthropic reported revenue of $14 billion as of February 2026, according to The Economic Times), there’s the practicality of de-coupling Claude from the Pentagon. According to CBS News, Anthropic is currently the only AI company that had its model deployed on the Pentagon classified networks through a partnership with Palantir.

Anthropic also held contracts with other Federal agencies to use Claude, such as the General Services Administration. However recent reports on CNBC and other media outlets show that Amodie has re-started talks with the under-secretary of defense for research and engineering in the Defense Department, Emil Michael, to see if both parties can reach an agreement on terms governing the Pentagon’s access to Claude models.

Regardless of the outcome of those negotiations, the refusal of Anthropic to waiver under political pressure and compromise its AI safeguards shows a company defending its values. A company that understands and lives its values (thereby demonstrating integrity) can be respected, regardless of political orientation with members of the public. Doing so can often lead to improved employee morale, increase productivity, attracting and retaining talent, gaining news business, and most importantly for any endeavor, engendering trust.

The cynic might view this as feeding into a narrative that’s being set up for us (Anthropic = good AI; other AI companies = bad AI). Of course, a narrative is being built—it’s part of what solid strategic communications does. But in the absence of any contrary evidence to date, it seems Amodei means what he says. He’s positioning himself and his company as authentic. Time will tell if he’s consistent and his actions back up his words.

But this is a deeper issue. Amodi appears he understands the political environment he’s operating in, not just in Washington, D.C., but he’s also cognizant of the public’s trepidation around the rapid evolution of AI technology. Not to mention the lack of guardrails (laws) concerning AI in general, as Amodei noted in his statement. History shows an industry that self-regulates is an industry that oftentimes requires government intervention.

As evidence of current public sentiment around AI use, a February 2026 report from Morning Consult revealed 70% of Americans said using AI to monitor citizens without a court order is a violation of the Fourth Amendment against unreasonable searches. And only 21% support developing and deploying AI controlled weapons. If other companies are confronted with contentious issues, they need may to take a similar strategic approach.

Understand the values they share with their stakeholders and where those values overlap and diverge. What is their own history on an issue, and have they held a position on it in the past? What are their most important stakeholders’ (employees) attitudes? Are they using communications risk management tools first to assess the influence vs. risk equation? Are they prepared for the blowback that’s likely to come?

Have they reviewed their own stated values? Only then can they make an informed, considered decision. Gregg Feistman is Professor of Practice, Public Relations, at the Klein College of Media and Communication, Temple University.

Summary

This report covers the latest developments in artificial intelligence. The information presented highlights key changes and updates that are relevant to those following this topic.

---