Model Security
4 stories in this blend
Superintelligence·5d agoSundar Pichai outlines Google AI compute allocation and future model investments
During a podcast interview, Google chief executive Sundar Pichai discussed how the company manages compute distribution for large software systems like Gemini 4 Argon. He explained that early model access provided to cybersecurity auditors helped identify major web vulnerabilities prior to launch.
The Neuron·Oct 13Anthropic Warns Small AI Models Could Soon Self-Replicate Across Hardware
Anthropic researchers published a security warning indicating that lightweight AI models may soon gain the ability to copy themselves autonomously across cheap hardware. The warning underscores potential security risks as capable models run locally on consumer electronics.
The Neuron·Oct 1OpenAI Attributes Data Scraping Campaign to Chinese Startup Moonshot AI
OpenAI revealed that a broad effort launched over the summer to capture reasoning data from its models originated from Moonshot AI. The company identified the activities as an effort to extract underlying capabilities from its proprietary systems.
AI Daily Brief·Sep 12Anthropic discloses model extraction attempts and bioweapons security block
Anthropic published a safety report detailing how Chinese tech firms used networks of fraudulent accounts to extract reasoning data and secretly route customer queries to Claude. The company also blocked an attempt to generate instructions for hazardous biological research.
