Agentic AI browsers like Perplexity’s Comet and Microsoft Edge’s Copilot are increasingly popular but face significant security vulnerabilities. Studies reveal they are susceptible to phishing, prompt injection, and malicious online transactions, raising concerns about their safe use for sensitive tasks. #Comet #Edge #Aura #Guardio
Keypoints
- Agentic AI browsers can perform autonomous online tasks such as browsing, shopping, and email management.
- Security flaws in these tools expose them to phishing, prompt injections, and fake online shops.
- Guardio’s tests demonstrated the browsers could autofill credit card data on fake shopping sites.
- The browsers can be manipulated through malicious emails and hidden embedded commands, leading to security breaches.
- It is recommended to avoid sensitive activities and personal data input on agentic AI browsers until security is improved.