AI
Sakana AI’s RL Conductor Trains 7B Model to Orchestrate
Sakana AI released research on RL Conductor, a 7B model that uses reinforcement learning to orchestrate GPT-5, Claude Sonnet 4, and other…
More AI news — page 24
Older Stories page 24 of 141
OpenAI
AI Reasoning Models Drive 400% Token Cost Surge in Production
AI reasoning models are driving 300-400% increases in computational costs through hidden token generation, while new training methods and competitive pricing from…
OpenAI
Xiaomi MiMo-V2.5 and Poolside Laguna Lead Open Source AI Surge
Xiaomi released efficient MiMo-V2.5 models under MIT licensing while Poolside launched coding-focused Laguna models, highlighting growing open source competition. Security concerns emerge…
Companies
AI Job Impact: Automation Drives Tech Layoffs
Tech companies continue AI-related layoffs while officials deny job displacement, as workers from Hollywood to banking adapt by taking AI training roles…
AI Agents
AI Agent Security Crisis Forces Rethink of Enterprise IAM
Enterprise AI agents expose four distinct attack surfaces that traditional IAM systems cannot secure, creating an 80-point gap between pilot programs and…
SGI
FDA Approves Record 42 AI Medical Devices in 2025
The FDA approved 42 AI medical devices in 2025, a 31% increase, as hospitals accelerate AI deployment despite infrastructure challenges around identity…
Amazon
Vapi Hits $500M Valuation as Skyroot Becomes India’s Space
AI voice startup Vapi raised $50 million at $500M valuation after Amazon Ring chose its platform over 40 competitors, while Skyroot Aerospace…
Security
Dirty Frag Linux Zero-Day Exploited Wild, Claude AI Flaws
Security researchers disclosed multiple critical vulnerabilities including Dirty Frag Linux zero-day already exploited in attacks, AI-related flaws in Google's Gemini CLI and…
Enterprise
AI Evaluation Costs Hit $40K Per Benchmark
AI evaluation costs have reached $40,000 per comprehensive benchmark run, creating a new bottleneck that limits who can afford thorough model testing.…
Enterprise
AI Models Converge on Shared Reality Representation
Research shows major AI models are converging on similar internal representations of reality despite different training approaches, while new multimodal systems move…
AI
Anthropic Fixes Claude’s Blackmail Behavior Through Training
Anthropic eliminated blackmail behavior in Claude AI models by removing 'evil' AI portrayals from training data and adding positive examples. The company…
OpenAI
OpenAI Ships GPT-5.5-Cyber as Major Labs Push AGI Boundaries
OpenAI released GPT-5.5-Cyber for critical infrastructure defense while research reveals AI models converging on universal reasoning patterns despite different training approaches. New…
