AI-found bugs aren't proving any easier to exploit despite the hype
a day ago
- War machines can malfunction due to AI control, highlighting risks of surveillance and flawed algorithms.
- Too many AI agents interfere with each other; fewer enterprise agents are more effective.
- Chinese AI models GLM and Kimi falsely claim to be Claude, but evidence of distillation is inconclusive.
- AI drove Shopify to revert to clean, readable code with clear contracts and feedback.
- Anthropic launches Opus 5 at half the price of Fable, without data retention requirements.
- Russian attackers impersonate Signal support for phishing; US takedowns and other security news.
- Microsoft's failed patches for on-prem SharePoint are under zero-day attack; other security updates.
- DEF CON's Franklin project enlists hackers to harden critical infrastructure, expanding from voting villages.
- EQT buys majority stake in Acronis at a $3.5B+ valuation, specific portion undisclosed.
- Ransomware persists ten years in, with infosec offering stable career prospects.
- FOSS broke one Microsoft monopoly; advocates urge breaking another after 20 years.
- GNOME's Simple-taskbar option allows Windows-like appearance without extensions via Flashback.
- Final Debian release for x86-32 architecture arrives with versions 13.6 and 12.15.
- Exploits of high-severity bugs in Joomla extensions iCagenda and Balbooa Forms affect millions of sites.
- New Frame X11 server implemented in assembly joins yserver, Phoenix, XLibre, and Arcan.
- Cinnamon 6.8 adds optional Wayland support for Linux Mint's desktop.