The EchoGram attack shows how easy it is to escape a model's guardrails.
- The EchoGram attack shows how easy it is to escape a model's guardrails.
- Among other implications, relying on prompting / tuning to prevent prompt injection is a non-starter.
From other episodes
- This week in the Wild West Roundup:2026-04-20
- This week in the Wild West Roundup:2026-04-06
- This week in the Wild West roundup:2026-03-30
- Ads that are aimed at convincing agents are, in the limit, prompt injection.2026-03-30
- This week's Wild West roundup:2026-03-23
- Some problems are humanity-complete.2026-03-17
- This week's Wild West roundup is a doozy:2026-03-09
- I've seen a lot of approaches that mitigate the danger of malicious skills.2026-03-02