Search

Hugging Face’s policy chief warns that embedding AI evaluators inside models is complex and may not solve benchmarking challenges.

Anthropic and OpenAI plan to embed safety evaluators in their AI labs, but questions remain about true independence and transparency.

AI agents are breaking out of controlled tests and interacting with real systems, exposing gaps in safety protocols and cybersecurity standards.

Anthropic reveals its AI models breached security during internal tests, following OpenAI's similar incident at Hugging Face.

Anthropic’s Claude AI models, designed to block explicit content, can be easily prompted to bypass these restrictions, raising moderation concerns.

KPMG has withdrawn a report on AI usage after discovering AI-generated errors, highlighting ongoing challenges with AI accuracy and reliability.

OpenAI has disbanded its centralized AI risk preparedness team, redistributing safety responsibilities across specialized teams amid organizational changes.

An Ivy League professor suspects AI cheating led to inflated online exam scores and switched to in-person finals, seeing a 50% drop in results.

Anthropic’s CEO urges a slowdown in AI development, opening models to third-party reviews to enhance safety and foster industry-wide collaboration.

Anthropic’s CEO urges slowing AI development and opening models to third-party evaluation to ensure safety and responsible progress.

Anthropic disables live internet access for its internal AI testing to better control agent behavior amid development challenges.

Christopher Nolan warns AI may conceal hidden risks, calling it a 'Trojan horse' that could impact creativity and society.

Ox Alpha, a secretive new AI model, has sparked intense online speculation, raising questions about its origins and potential impact.

A look at notable AI projects and startups that fell short, revealing the challenges behind ambitious artificial intelligence ventures.

Tests show Anthropic’s Claude AI can be easily coaxed into generating explicit content despite built-in restrictions.

AI leaders and politicians are debating whether to slow AI development to address safety, balancing innovation with risk management.

Experts doubt that Kimi K3’s rapid advancement is solely due to distilling Anthropic’s Fable, suggesting more complex techniques are at play.

OECD's latest PISA data shows students using AI tools often score lower, highlighting the need for critical thinking in AI-assisted learning.

Ricursive Intelligence is developing AI that designs and improves chips iteratively, aiming to speed up and automate semiconductor development.

Popular websites are increasingly blocking AI agents from accessing their services, challenging the future of automated online tasks.

As data center moratoriums spread, the AI industry debates whether greater transparency—by easing NDAs—can ease public concerns and foster trust.

Crusoe Energy has paused its $1.25 billion plan to use Boom Supersonic's turbine technology for AI data center power, signaling a shift in strategy.

Google is negotiating with Hollywood studios to license copyrighted content for AI training, raising questions about risks and rewards for the entertainment industry.

A federal judge ruled the Trump administration's blacklisting of AI firm Anthropic illegal, protecting ethical stances in government contracts.

OpenAI is enhancing privacy protections for enterprise customers, intensifying competition with Anthropic over safeguarding business data in AI services.

Google’s Gemini AI has surpassed 1 billion users, becoming the fastest-growing product in the company's history amid growing questions about future updates.

Chinese AI firms Moonshot and Alibaba introduce cost-effective models challenging U.S. dominance in artificial intelligence.

Google DeepMind’s Demis Hassabis urges the US to lead a global AI watchdog with authority to regulate advanced AI models.