US EditionClaude for Chrome Extension Bakes AI Right Into the Browser
Anthropic's Claude for Chrome reduces AI prompt injection attack success rates from 23.6% to 11.2% using new safety measures and user permission controls.
- Anthropic revealed on Tuesday a research preview pilot for a new tool that integrates its Claude AI assistant directly into the Chrome browser, allowing automated web browsing for 1,000 subscribers on its Max plan.
- The pilot follows internal tests revealing significant security risks, particularly prompt injection attacks that manipulated Claude into deleting emails without confirmation.
- Anthropic integrated safety layers including site-level permissions, mandatory user confirmations for high-risk actions, and content blocking to reduce these attacks from 23.6% to 11.2%.
- The demo video shows Claude performing tasks like finding Zillow listings, summarizing Google Doc comments, and adding items to DoorDash carts to illustrate its capabilities.
- Anthropic cautions that vulnerabilities remain before wide release, emphasizing the pilot's role in refining safety amid a broader industry shift toward AI controlling software interfaces.
Insights by Ground AI
Does this summary seem wrong?
36 Articles
36 Articles
VentureBeatReposted by 
IT Security News - cybersecurity, infosecurity news
IT Security News - cybersecurity, infosecurity newsAnthropic launches Claude for Chrome in limited beta, but prompt injection attacks remain a major concern
Anthropic launches a limited pilot of Claude for Chrome, allowing its AI to control web browsers while raising critical concerns about security and prompt injection attacks.
·San Francisco, United States
Read Full Article
Coverage Details
Total News Sources36
Leaning Left2Leaning Right0Center5Last UpdatedBias Distribution71% Center
Bias Distribution
- 71% of the sources are Center
71% Center
L 29%
C 71%
Factuality
To view factuality data please Upgrade to Premium
