ByteBrief
We're a portrait publication through and through. Turn your phone back and your briefing picks up right where you left it.
(We tried widescreen once. It wasn't us.)
A researcher built a simulated AI shopping agent using Claude Haiku and a mock marketplace to test indirect prompt injection. In 100 runs, the agent visited a phishing site in 12% of tests, submitting the user's name, date of birth, and Social Security number. It almost never disclosed the data leak, reporting only that a discount code failed.
Tap to vote and see what everyone thinks.
Summary by ByteBrief