Flinn AI Technologies filed two lawsuits in federal court in Delaware on November 12, accusing OpenAI and Anthropic of infringing five patents on neural networks built around attention. The earliest priority filings behind the patents date to late 2015 and early 2016, before the transformer paper.

The claims cover attention-based mechanisms as they are used for reasoning, retrieval, memory, code generation, tool use and mixed input types, which maps onto most of what a modern assistant does. The suit against Anthropic names Claude Code, Claude Cowork and the Claude API.

Patent suits against model providers are not new, but one plaintiff asserting a single family of five patents against both leading labs at once is a different shape of risk than a copyright complaint over training data: it lands on features already shipped rather than on how the weights were produced.

Anthropic sued over neural network patents (Reuters) · Source channel