AI model watermarking changes agent behavior
4 hours ago
- AI model watermarking changes agent behavior, highlighting differences in tool handling and model refusals.
- Nvidia goes green to manage grid capacity and protect revenues.
- Spain experiences its first AI-aided cyber attack, prompting calls for review of data protection models.
- AI networking startups like Cornelis and Delos Data race to replace Nvidia's NVLink for scaling AI.
- Anthropic and OpenAI seek government backing to cement their dominance in the AI market.
- Russians pose as Signal support for phishing attacks; US takes down Iranian propaganda sites.
- Microsoft's failed patches leave on-prem SharePoint under zero-day attack.
- DEF CON's Franklin project enlists hackers to harden critical infrastructure.
- EQT buys majority stake in Swiss cybersecurity firm Acronis at a $3.5+ billion valuation.
- Ten years since first corporate ransomware, Mikko Hyppönen sees no end to the threat.
- Shopify acquires Tailwind CSS, providing a stable open-source home for the framework.
- Switzerland tests a FOSS escape route from Microsoft 365.
- Kumander Linux offers a modern desktop experience based on Debian and Xfce.
- Canonical shuts down legacy chat channels and demotes IRC.
- Audacity audio-editing app gets a fresh UI and new features.
- Haiku OS releases Beta 6, offering an alternative to FOSS Unix clones.