Web & Browser Security: AI Uncovers Zero-Days, Critical Sandbox Escapes

Anthropic's Claude Mythos Uncovers Thousands of Zero-Day Vulnerabilities

• Anthropic's new AI model, Claude Mythos, has demonstrated the ability to autonomously identify and exploit thousands of zero-day vulnerabilities across major operating systems, enterprise applications, and web browsers.
• This AI capability extends beyond vulnerability identification to include exploitation, showcasing a paradigm shift towards proactive vulnerability discovery and defense, branded as Project Glasswing.
• The development highlights the increasing role of advanced AI in cybersecurity, aiming to accelerate the discovery and remediation of critical flaws before they can be leveraged by malicious actors.

Source: Help Net Security | Date: April 08, 2026

Zero-Day Fingerprinting Attack Targets Adobe Reader Users

• Security researchers have identified an active zero-day attack specifically targeting Adobe Reader users through a sophisticated fingerprinting technique.
• This attack vector leverages specific PDF rendering behaviors to gather detailed system configuration information, enabling the delivery of highly tailored exploits.
• Users are advised to exercise extreme caution with untrusted PDF files and ensure all software is up-to-date, as no patch is yet available for this actively exploited vulnerability.

Source: The Cyber Express | Date: April 8, 2026

References

1. Anthropic's Claude Mythos Finds Thousands of Zero-Day Flaws Across Major Systems - The Hacker News
2. Anthropic’s new AI model finds and exploits zero-days across every major OS and browser - Help Net Security
3. Project Glasswing powered by Claude Mythos: defending software before hackers do - Security Affairs
4. Researchers Find a Zero-Day Attack Targeting Adobe Reader Users - The Cyber Express