Safety
U.S. Military Halts Operation Due to AI Hallucination
A U.S. military operation against a Chinese vessel was aborted after AI-generated intelligence falsely claimed the ship carried nuclear components, according to CNN.
LATEST NEWS
-
World Model Companies Keep Secrets Amid Uncertain Commercialization
World model developers like AMI Labs and World Labs are reluctant to share details about their products, despite the field's potential for lucrative applications in robotics and media.
-
Anthropic Introduces Accenture as First Embedded Evaluator
Anthropic announced it will embed Accenture’s evaluators in its AI labs, investing $1 billion over five years to enhance model safety.
-
42 Mathematicians Warn of AI Existential Risk
42 leading mathematicians, including Fields Medalists, warn that advanced AI poses an existential risk exceeding 10%.
-
Anthropic Reveals Claude Leads 26% of Its Research
Anthropic disclosed that Claude leads 26% of its research work, but the term 'lead' does not imply full autonomy, according to the company.
-
Deepmind Warns Transparency in AI Chains of Thought is at Risk
Deepmind researchers say visible chains of thought in AI models like Gemini 3 Pro are a safety advantage, but transparency is declining as newer models become harder to monitor.
-
Media Companies Call OpenAI's AI Training 'Astonishing Theft'
The New York Times and other media firms are suing OpenAI and Microsoft, seeking billions in damages over alleged copyright violations in AI training.
-
California Demands Kill Switch for AI Models
California Governor Gavin Newsom signed an executive order requiring AI companies to embed independent auditors and implement kill switches, following the Hugging Face attack.
-
AI Safety Roundtable Explores Risks and Alignment
A live roundtable event hosted by MIT Technology Review examined AI risks, with experts discussing potential dangers and alignment challenges.
-
Left Divided Over AI Safety, With Sanders Leading the Charge
Democratic Socialists of America posted a critical take on AI alarmism, sparking debate over whether to prioritize theoretical risks or current harms, with over 25,000 Instagram likes.
-
Basura and Big Data Launch Data Center Plushie
Basura and Big Data released the Bezzy plushie, a satirical data center mascot, on April 5, 2024, with a price tag of $99.
-
Anthropic CEO Warns of AI Risks Despite Industry Momentum
Anthropic CEO Dario Amodei admits the industry is largely in the dark about why AI models sometimes act unpredictably, citing a 10% chance of human extinction.