Why could it take months to replace Anthropic at the Pentagon?

by
0 comments
Why could it take months to replace Anthropic at the Pentagon?

How exactly does the Pentagon eject the cloud?

It takes minutes to swap one AI model for another on a classified network. People who have learned to trust it will take longer to retrain.

An aerial shot of the Pentagon

The Defense Department is phasing Anthropic’s cloud out of its classified networks within six months, triggering a complex transition for military personnel.

AFP/Stringer/Getty Images

The Pentagon has put Anthropic on watch. On Thursday, the Defense Department formally notified the company that it has been deemed a “supply chain risk” — a label that has turned artificial intelligence systems, including its flagship model, the cloud, into a liability.

The move escalates the controversy that has been ongoing for several weeks regarding Anthropic safety-first ethos– Its commitment to limit how its technology can be deployed – and the DOD’s demand for unfettered control.

Pentagon is Cloud phasing outOne of the world’s most advanced AI models, from its classified network within six months. On paper, replacing one model with another appears quick. According to a source close to defense-tech giant Palantir, “It is easier to replace models and install new ones.” Partnered with Anthropic Cloud hosting inside a secure military network.


On supporting science journalism

If you enjoyed this article, consider supporting our award-winning journalism Subscribing By purchasing a subscription, you are helping ensure a future of impactful stories about the discoveries and ideas shaping our world today.


The hardest part starts after the model is finished, reassembling everything built around it.

The cloud is known as the Frontier Model, an AI capable of performing complex, multistep tasks on its own. DoD does not currently use it in this manner. Lauren KahnA researcher at Georgetown University’s Center for Security and Emerging Technology and a former Pentagon official describes its deployment as more like a chatbot than a free-roaming agent. The cloud sits “on top” of existing software, she says, and is visible only in certain places — in tightly controlled corners of classified environments. And it’s not linked to “effectors,” she says, meaning it can’t “launch an effect” — a weapons command, for example — “in the real world.”

At the end of 2024, Anthropic becomes the first AI company Clear Pentagon classified barriers. Until recently, the cloud was the only major language model Publicly known for working in that environment. Access through tools such as Cloud Gov – which became a preferred option for some defense personnel, According According to Bloomberg – The system uses vast data pipelines to transform floods of unstructured information into readable intelligence. In other words, the cloud summarizes information for the Defense Department, but it can’t pull any triggers.

Once people rely on a device, it can be hard to abandon it. Each integration must be removed piece by piece. And whoever replaces the cloud will have to complete rigorous security reviews and approvals before touching any classified systems. Software changes inside the Pentagon can be “painful,” Kahn says. Even something as simple as installing Microsoft Office “took months and months and months.”

As of press time, Anthropic did not respond to multiple requests for comment. scientific American. The Defense Department declined to discuss the specifics of the infection.

unlearning the cloud

Every AI model fails in its own specific ways. Operators who have spent many months using the cloud learn those quirks through trial and error: which signals land badly, which ones require output to be recalibrated.

Kahn studies automation bias, the tendency for human operators to assign more work than necessary to machines. “I’m concerned about a slightly increased risk of automation bias in the early stages as they’re working on the bugs,” she says. People will check out the cloud’s mistakes while the replacement model makes new mistakes. The personnel most exposed to the change will be the power users who created the most optimized work flows and learned the shortcomings of the model well enough to exploit its strengths.

As Pentagon personnel prepare for operational changes, messy details of the political standoff have emerged to the public. Published late Thursday night by Anthropic CEO Dario Amodei a blog post Vowing to challenge the government’s “supply chain risk” designation in court, arguing that the statute is typically reserved for foreign adversaries. It appears that behind the scenes the standoff has turned into a game of chicken. Emil Michael, the Pentagon official who led the department’s negotiations with Anthropic, Posted on x Negotiations with the company have stopped. and amodei is alleged scuffle To revive them.

Meanwhile the DOD is already moving forward. Within hours of Anthropic’s official blacklisting, OpenAI announced It had signed a deal to deploy its models on the military’s classified networks, securing the contract lost to its rival.

Anthropic was willing to risk expulsion from the US government rather than compromise its security-first ethos. Its replacement initially merely met the Pentagon’s demand for uninterrupted operational flexibility. add quickly The same surveillance guardrails that Anthropic advocated for after OpenAI CEO Sam Altman faced massive internal and public backlash. The swap may not be that simple after all.

It’s time to stand up for science

If you enjoyed this article, I would like to ask for your support. scientific American He has served as an advocate for science and industry for 180 years, and right now may be the most important moment in that two-century history.

i have been one scientific American I’ve been a member since I was 12, and it’s helped shape the way I see the world. Science Always educates and delights me, and inspires a sense of awe for our vast, beautiful universe. I hope it does the same for you.

if you agree scientific AmericanYou help ensure that our coverage focuses on meaningful research and discovery; We have the resources to report on decisions that put laboratories across America at risk; And that we support both emerging and working scientists at a time when the value of science is too often recognised.

In return, you get the news you need, Captivating podcasts, great infographics, Don’t miss the newsletter, be sure to watch the video, Challenging games, and the best writing and reporting from the world of science. you can even Gift a membership to someone.

There has never been a more important time for us to stand up and show why science matters. I hope you will support us in that mission.

Related Articles

Leave a Comment