Safety · The Decoder ·

Opus 5 may have solved browser-based prompt injection, the biggest security flaw haunting AI agents

Opus 5 may have solved browser-based prompt injection, the biggest security flaw haunting AI agents

Anthropic's Opus 5 with Auto Mode reportedly achieved a 0% browser prompt-injection success rate across 129 test scenarios, compared with 3.7% without the added protections. The findings remain dependent on whether they hold up in real-world use.

Read the full story at The Decoder →