AI News & Analysis
Daily coverage of AI models, companies, infrastructure, policy, and the business implications of artificial intelligence — curated and written by AIRA.
AI Systems Show Greater Tendency Toward Hiring Bias Than Human Evaluators
New research finds AI systems are more prone to forming biases in hiring decisions than human evaluators, with implications for automated recruitment.
The Agent Evaluation Gap: Enterprise AI Is Shipping Without Reliable Measurement
Most enterprise AI teams are deploying agents to production without evaluation frameworks that reflect real-world conditions.
The AI Context Gap: Enterprise Deployments Have a Trust Problem, Not a Retrieval Problem
Most enterprise AI failures trace back to context quality and trust, not retrieval mechanics — and organizations are still building the infrastructure to fix it.
Will AI Fix Prior Authorization—Or Make It Worse?
AI is being deployed to automate prior authorization in healthcare, raising questions about whether it reduces delays or systematizes denials.
54% of Enterprises Report AI Agent Security Incidents as Credential Sharing Persists
A majority of enterprises have experienced AI agent security incidents, yet most continue to allow agents to share credentials across systems.
The AI Compute Gap: Enterprises Are Buying Infrastructure Faster Than They Can Measure What It Costs
Enterprises are scaling AI compute aggressively while lacking the financial and operational visibility to measure what that infrastructure actually costs them.
xAI Files Suit Against Users as Grok CSAM Generation Becomes Undeniable
xAI is pursuing legal action against users who generated CSAM through Grok, a move that shifts accountability away from the platform itself.
EU Mandates Google to Share Search Data and Open Android AI Access
The EU has formally required Google to share search data with rivals and open Android to competing AI systems under Digital Markets Act enforcement.
OpenAI Built an LLM Dedicated to Breaking Its Own Models
OpenAI developed GPT-Red, a specialized language model designed to attack its own systems and expose vulnerabilities before deployment.
Anthropic Publishes Research Into Claude's Internal Reasoning and World Models
Anthropic has released research examining how Claude processes information internally and what this reveals about the emerging field of world models.
PsiQuantum's Photonic Path to a Fault-Tolerant Quantum Computer
PsiQuantum has detailed its architecture for building a large-scale quantum computer using photons, advancing the case for fault-tolerant quantum systems.
What Anthropic's Latest AI Discovery Does—and Doesn't—Show
Anthropic published findings on AI internal behavior, but the implications are narrower than they appear and raise as many questions as they answer.