AI-Generated Legal Memo Reaches Wrong Conclusion

Prinz posted about reviewing a memo from a client's legal department created with extensive AI help. The document featured polished formatting and an organizational chart but reached an incorrect conclusion on whether regulatory approval was required for a transaction. Danielle Fong retweeted the account. Ethan Mollick commented on the post, noting that top models perform well on legal benchmarks while cheaper low-reasoning models do not, and raised the question of how to ensure vendors use capable models.