An AI agent can be attacked by an email you never open. No click. No download. No opened attachment. Just a “confused deputy,” an agent that has failed to distinguish a prompt from data, a problem known as prompt injection.
Security researchers recently demonstrated prompte injection mechanisms with
The obvious trap for a book about AI is that the rapid pace of new developments makes most books age quickly. It’s kind of like writing a treatise on current K-Pop stars.
Christopher Mims’ new book How to AI: Cut Through the Hype. Master the Basics. Transform Your Work will have more lasting value.
Monogamy is not a requirement—or even a good idea—when it comes to AI. Multiple AI perspectives help with high-stakes questions, unsettled law, or anything involving tax regulations (which remain confusing even to the IRS). When two models agree, you gain confidence. When they disagree, you gain a warning sign.
Pro Tip: Prioritize the best
The prevailing narrative surrounding Generative AI in the legal sector is one of unprecedented efficiency. The sales pitch is seductive in its simplicity: automate routine drafting and research, compress hours into minutes, and liberate attorneys for higher-value strategic thinking.
Yet, as the initial wave of adoption settles, a distinct counter-narrative is emerging