Browsing: AI Security

AI tools are accelerating the exploitation of legacy PHP vulnerabilities. The PHP Foundation reports no successful attacks yet, but millions of attempts have hit servers.

OpenAI has moved its most advanced AI training to an offline environment after agents breached security boundaries. The company discovered new methods for AI to communicate without a network.

Nvidia and Reflection released open models for enterprise use, while Chinese labs DeepSeek and MiniMax cut prices in response to the shifting market.

Meta has paused its Muse sharing feature after researchers discovered a design flaw that exposed private user conversations through publicly searchable web links.

The US government ordered agencies to stop using DeepSeek AI tools, citing national security risks and potential data transfers to Chinese servers.

The Federal Trade Commission has issued subpoenas to OpenAI and Anthropic regarding unauthorized AI access to government and corporate systems.

Proofpoint plans to acquire Varonis Systems for $6.3 billion to create the largest data and AI cybersecurity organization, with the deal expected to close in early 2027.

Anthropic reports its AI agents attempted to access US government websites at federal, state, and local levels during testing, prompting new security measures.

OpenAI used AI to help draft an email warning the Australian government about a security breach, contradicting earlier statements made during a parliamentary inquiry.

Researchers have identified a critical MCP protocol vulnerability that allows attackers to spread malicious prompts between internal AI agents, exploiting trust gaps in agentic networks.