Last week, a contract negotiation between the Department of War (DoW) and Anthropic turned into a forced vendor switch across parts of the federal government. Treasury and the Federal Housing Finance Agency said they are ending use of Anthropic products. The takeaway for small and medium-sized businesses is a lesson in the importance of avoiding vendor lock-in.
In case you missed it, the DoW wanted to ensure it could use Anthropic’s models for “any lawful use,” which could include autonomous weapons and mass surveillance. Anthropic refused, which resulted in the government cutting ties with the AI lab and labeling it a supply-chain risk. Digging just a little bit deeper, we find that Anthropic’s reasoning was less noble than it might seem: In short, the company thinks its systems are simply not yet capable of autonomous weapons, and thus deploying them would be a risk to U.S. warfighters. Their stance on mass surveillance is that it would violate American civil liberties.
The legal and moral questions the episode raised are interesting, but not really the point for this newsletter. Instead, it is a great reminder that building an entire ecosystem around a single model is a huge risk. Federal agencies have six months to transition away from Anthropic. In the private sector, this should be doable in a week or less.
There are plenty of tools to avoid this situation for your own business. For instance, Langchain and other open source libraries make it very easy to abstract away the slight differences between AI APIs, allowing developers to quickly “drop in” a new model across an entire project with just one line. N8N (an automation framework everyone should be learning) workflows are similarly simple to update with new models and rapidly test with evaluations. And if you have many workflows that need updating, Claude Code can interface directly with N8N to do it for you.
These are simple principles to apply, but easy to forgo if you don’t have someone in your team that has already learned this lesson.
If you found this newsletter useful, forward it to someone who would too. Click here to learn how Applied.AI can help your business make AI work for you.
