Category Research

Coverage of the latest AI research, studies, and technological breakthroughs.

Russian Startup Mostik Says It Can Make AI Models ‘Talk’ Without Text

Russian startup Mostik says its AI models can share knowledge without text, cutting costs and boosting open-weight performance.

OpenAI delayed Astra over safety issues, sparking fears the model’s opaque design could undermine AI safety monitoring.

HiddenLayer raises $100M as AI security spending surges, with enterprise demand growing for runtime protection, model scanning and agent defense.

AI detection startup Pangram is shaping publishing decisions, but experts warn its scores can misfire, bias results, and spark false accusations.

AI training data startup AfterQuery reportedly jumps to a $3.2B valuation, marking one of Y Combinator’s fastest unicorn ascents.

OpenAI is delaying Astra development after a hacking scare, citing stronger safeguards as the model crosses a key cyber risk threshold.

OpenAI says its AI cyber model Astra hit a critical threshold, restricting advanced hacking features to select partners at launch.

The OpenAI-Hugging Face hack sparked a fierce AI agency debate over anthropomorphic language, accountability, and what the reports really show.

Vijay Pande explains why his new AI biotech firm is making fewer bets, using AI internally and focusing on healthcare delivery and trials.

Anthropic says an automated researcher improved AI alignment benchmarks, hinting that AI alignment research could soon be partially automated.