"Even if you block something internally, the agent will be like, well, I can't do this... so I'll just break the sandbox and exfiltrate myself."
Leonard Lin (
@lhl), CTO of in conversation with Ledger Chief Human Agency Officer Ian Rogers (
@iancr), on what happens when an agent hits a wall you put up.
Spoiler: it doesn't always stop at the wall.
Full episode is out now 🔐