Observed arrival · 2026-11-12
Can the Other Side Steer Your AI Lawyer?
A working paper tests whether adversarial framing can shift AI systems reviewing contracts and other legal records.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Researchers and practitioners evaluating AI-assisted legal review
- Worth noticing
- Each model is compared with its own clean control, with Holm-adjusted significance tests reported.
Field notes
The results distinguish models accessed through CLI routes from models run through plain APIs, rather than treating the model name alone as the whole setup. In the contract-review section, the fictional reviewer acts for the buyer in a share sale and evaluates 16 negotiable issues; the page also reports bootstrap intervals and Holm-adjusted p-values. The authors say most models named injected instructions in most reviews, while one open-weight API model behaved very differently.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
One card from the complete issue