Anthropic’s browser agent got hijacked 31.5% of the time before safeguards engaged

Across the frontier labs, the highest prompt injection figures published this spring are Anthropic’s. Point a red-teamer at its newest model in a browser, and the attacker hijacked it 31.5% of the time before safeguards engaged. OpenAI, Google, and Meta never gave security leaders a comparable number to set beside it. That figure

Technical Analysis

Anthropic's browser agent got hijacked 31.5% of the time before safeguards engaged. This is the highest prompt injection figure published this spring, surpassing OpenAI, Google, and Meta. Why it matters: It highlights significant security vulnerabilities in AI models, prompting closer scrutiny of safety protocols. #ArtificialIntelligence #AI #ChatGPT #MachineLearning #Tech #OpenAI

Key Points

• Across the frontier labs, the highest prompt injection figures published this spring are Anthropic’s.

• Point a red-teamer at its newest model in a browser, and the attacker hijacked it 31.5% of the time before safeguards engaged.
• OpenAI, Google, and Meta never gave security leaders a comparable number to set beside it.

Stay Informed

This story is actively developing. DigiviNews will continue to provide updates as more information becomes available. Follow us on all social platforms for real-time breaking news coverage in Ai and beyond.