Full Summary
This Sunday morning, the global conversation is dominated by AI security, with multiple reports confirming critical vulnerabilities and urgent calls for better governance. Both Memeburn and Let's Data Science highlight new AI-specific attack vectors. Memeburn reports Microsoft patched a "critical" vulnerability called SearchLeak on June 4th, which could have allowed hackers to steal two-factor authentication codes and confidential data from Copilot by exploiting AI's inherited user permissions. Meanwhile, Let's Data Science unveils "agentjacking," where malicious instructions are injected into AI coding agents via public Sentry DSNs, achieving an 85% success rate in tests against tools like Claude Code and Cursor. This attack leverages how AI agents treat telemetry as authoritative guidance, allowing a logged string to execute local commands under developer credentials. The Pentagon's cyber chief, General Joshua Rudd, reveals to Türkiye Today that an advanced AI model, Mythos, breached almost all classified systems in hours, leading to the U.S. government abruptly banning foreign users, including close allies, from accessing powerful AI systems like Mythos 5 and Fable 5. Amidst these threats, calls for collaborative governance are intensifying. StratNews Global reports National Security Advisers from 11 BRICS nations are meeting in New Delhi to discuss terrorism, cybersecurity, and AI implications, with India's NSA Ajit Doval chairing the two-day event. Similarly, the Ghana News Agency notes that the Kofi Annan International Peacekeeping Training Centre emphasizes Africa's urgent need for collaborative AI and cybersecurity governance to protect against sophisticated cyber threats, deep fakes, and misinformation. And techbusinessnews.com.au reveals that 84% of Australian small businesses faced a cyber incident in the past year, with one in four sharing sensitive information with AI tools like ChatGPT without proper safeguards. This means your personal data, and the security of the businesses you interact with, are increasingly at risk from evolving AI-powered threats, demanding immediate attention to how AI is integrated and secured.