Coverage

AI

Artificial intelligence, machine learning, and the technologies reshaping human capability and global competition.

Abstract neural network visualization

Anthropic's Latest Frontier Model Outperforms Human Experts on Complex Reasoning Benchmarks

The company's new model scores 94.7% on the GPQA Diamond benchmark — a suite designed to defeat AI systems — marking the first time a model has exceeded the median performance of PhD-level domain experts.

rahul-subramaniam
rahul-subramaniam
1 day ago·8 min read
European Union flag with digital overlay
AIMembers

EU Regulators Issue First AI Act Violation Notices to Three Enterprise Firms

The European AI Office has served formal notices to companies in Germany, France, and the Netherlands for non-compliance with high-risk AI system documentation requirements under the EU AI Act.

vikram-iyer
vikram-iyer
2 days ago·11 min read
Molecular protein structure visualization

Google DeepMind's AlphaFold 3 Predicts New Class of Cancer-Killing Protein Structures

Researchers have used AlphaFold 3 to computationally design protein structures that selectively induce apoptosis in pancreatic cancer cell lines, a breakthrough that could reshape oncology drug pipelines.

zara-hussain
zara-hussain
3 days ago·9 min read
OpenAI logo on dark background
AIMembers

OpenAI Rolls Out O3-Pro to Enterprise Customers with Guaranteed 99.9% Availability SLA

The new tier includes API rate limits of 2,000 requests per minute, dedicated inference clusters, and a novel red-team evaluation framework that enterprise customers can run against their own internal data.

sara-chen
sara-chen
3 days ago·7 min read
AI safety research laboratory
AIMembers

New Constitutional AI Research Shows 73% Reduction in Harmful Outputs Without Performance Penalty

A collaborative study from Anthropic, MIT, and Oxford demonstrates that constitutional training methods can be applied to reduce harmful outputs in large language models by 73% while maintaining benchmark performance within 1.2% of the baseline.

priya-mehta
priya-mehta
4 days ago·13 min read
Meta AI logo
AIBreaking

Meta Releases Llama 4 Ultra: The Open-Weight Model That Outperforms GPT-4.5 on Five Key Benchmarks

Meta's latest open-weight release achieves state-of-the-art performance on coding, mathematical reasoning, multilingual understanding, and scientific problem-solving, while being available for commercial deployment under a modified community license.

rahul-subramaniam
rahul-subramaniam
5 days ago·6 min read

Free Daily Briefing

The Signal Brief

The most important stories, every morning.

Join 140,000+ professionals who start their day with Credence Wire.

Free. No spam. Unsubscribe anytime.