Anthropic’s browser agent got hijacked 31.5% of the time before safeguards engaged
Source:
ventureBeat
June 1, 2026 · 08:50
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 looks like a liability. In this comparison, it is the opposite. It's the one solid piece of ground.Four frontier labs each shipped a prompt injection disclosure, and no two match. Anthropic put 244 pages and four agentic surfaces on the table on May 28. OpenAI reported one surface, connectors. Google moved the subject out of the model card and into a separate safety framework. Meta shipped no closed-model card at all. The Cross-Vendor Prompt Injection Disclosure G…
The original article opens on the publisher's website.
More from Automotive
View topic →Larry Page’s flying car company Pivotal loses its CEO
techcrunch
Sep 1, 2026 · 16:59
SEC proposes transfer agent rule, sets event to figure out round-the-clock U.S. trading
coinDesk
Sep 1, 2026 · 16:55
The Range Rover Electric: Specs, Price, Availability
wired
Sep 1, 2026 · 16:01
Android stuck in Safe Mode? Here's how to turn it off
engadget
Sep 1, 2026 · 16:00
How to change the background on iPhone Messages
engadget
Sep 1, 2026 · 15:30