The Anthropic-Pentagon Standoff And The No. 1 Lesson For Midmarket IT Leaders

The clash between Anthropic and the U.S. Department of Defense proves IT leaders must establish their own AI red line.

Anthropic’s refusal to remove safeguards from its AI models – even after direct pressure from the U.S. Department of Defense is being framed as a standoff between Big Tech and the federal government.

But for midmarket CIOs and CISOs, the real story isn’t about politics. Instead, it’s about establishing an AI red line.

What does an AI red line look like in practice? Pushback and governance are the core components of an AI red-line strategy.

First, it’s about pushing back on senior leadership’s AI-driven business goals that clash with technical realities, risk, and responsibility.

An AI red line also means creating strong policies and frameworks that clearly outline responsible use of AI within your organization, even though you may not have industry, local or federal AI regulation guidelines to fall back on.

Pushback Is An IT Leader’s Job

In July 2025, the DoD awarded Anthropic a $200 million contract to provide AI tools for national security applications.

For a while there it seemed to be a mutually beneficial relationship. Anthropic CEO Dario Amodei in a February 26 blog post ticked off several accomplishments Anthropic said it has made with the DoD contract: first frontier AI company to deploy its models in the U.S. government’s classified networks, and Claude “extensively deployed” across the DoD.

Then in January, the relationship began to break down as the DoD claimed it had the right to use Claude for “all lawful purposes.”

In response, Anthropic set its red line. The company refused the DoD’s request for unfettered use of Anthropic’s AI by adhering to its policy that includes guardrails against using its AI for mass surveillance and for creating autonomous lethal weapons.

Defense Secretary Pete Hegseth then said that Anthropic would be designated a “supply chain risk,” after not meeting the administration’s demands, and a label usually reserved for technology from nation-state adversaries.

Anthropic had to push back against the U.S. military. While perhaps not as high profile a battle, midmarket leaders also face pressure to implement AI posthaste, be it by a board, their CEO, or other stakeholders.

Which leads to the question IT leaders are quietly asking: How do you manage senior leadership’s fervor for AI as business transformation tool, to outmaneuver competitors, or to slash operational expenditure while conveying expectations and risk?

Just saying no will not work. Pushback doesn’t mean squashing leadership’s ambitions and goals around AI. Rather, it’s an opportunity to redirect leadership’s enthusiasm for AI into conversations about realistic goals that won’t pose reputation, financial, or legal risk to the organization.

There are three considerations for IT leaders when communicating with leadership about AI’s capabilities:

Where can AI be used as an adviser but not a decider?
When is human-in-the-loop non-negotiable?
What risks can AI bring that the organization absolutely cannot afford being exposed to?

Safe to say, most senior leaders value authenticity and transparency when discussing AI and expectations.

Governance Is The Second Half Of The AI Red-Line Strategy

Anthropic didn’t push back against the government’s demand by taking an anti-war or other ideological stance. It stuck to its already-established policy and governance rules despite immense, high-profile pressure. This is an important takeaway for midmarket IT. Before deploying any AI projects, rock-solid AI governance and policies are mandatory. IT leaders cannot wait for government or industry regulation but must have strong AI governance frameworks in place.

Here are some key steps to take when creating an AI governance strategy:

Get data governance implemented first. “AI governance is naturally an extension of data governance,” said Blaine Carter, FranklinCovey global chief information officer, during an appearance on the Ready.Set.Midmarket! podcast.

“If you and your organization didn’t have a great data governance program, chances are AI governance is going to be a very difficult task because AI is basically just making data accessible in a way that’s much more humanly consumable,” Carter said.

Get other departments involved. AI governance is not “an only-IT responsibility,” said Jay Ferro, chief information, technology and product officer at clinical trial management company Clario.

Instead, it’s an “organizational lift,” Ferro said. In his organization, department heads met to form a consensus on several AI governance topics: What is AI allowed to do? What data can AI access? Who is accountable for AI?

Get an AI policy in place. For starters, get a reasonable use policy in place. “You don’t have to get into some sort of arcane 80-page manual right out of the gate,” Ferro said. “Start with simple tiering,” define what in the organization is low, medium, or high risk. “Don’t let perfection be what’s stopping you from getting started,” Carter said.