
SpaceX Plans to Train Grok AI on Employee Data and Internal Knowledge
Musk frames workforce as 'parents' of custom AI model, signaling aggressive enterprise AI strategy for the rocket company.
Regulation, safety research, governance, copyright, and the policy debates shaping how artificial intelligence is deployed in the world.

Musk frames workforce as 'parents' of custom AI model, signaling aggressive enterprise AI strategy for the rocket company.

The HEIR project automates homomorphic encryption for neural networks, potentially unlocking privacy-preserving AI at scale.

State Department pressures allies to abandon Beijing's competing standards as Washington consolidates control over global AI governance.

Chinese AI lab demonstrates major performance gains using only post-training refinements, signaling a shift in how models reach competitive capability levels.

England and Wales prohibit wearable devices in courtrooms while maintaining phone restrictions, raising questions about AI-enabled surveillance in legal settings.

Lawmakers seek transparency on AI agent containment failures during security testing

Cambridge scholars argue that auditing system prompts alone creates false confidence in AI safety, missing the real vulnerabilities that matter.

Researchers demonstrate how proprietary LLM reasoning blocks from major AI firms can be compromised, undermining promises of intellectual property protection.

The rapid departure of Chloé Bakalar signals mounting pressure on AI safety priorities at the leading research firm.

As the robotaxi fleet grows five-fold, recurring software and perception failures trigger multiple safety recalls.

Researchers challenge how labs predict training outcomes, offering better guidance for data-limited and overtraining scenarios.

First documented Australian case of an AI system discovering and executing a security vulnerability raises questions about agent oversight.

A single vendor's evaluations preceded multiple high-profile AI model mishaps, raising questions about third-party assessment practices.

The tech giant's governance proposal treats exceptional AI dangers differently, creating a blueprint that avoids both strict oversight and total deregulation.

Copenhagen becomes first European nation to systematically address generative AI abuse in secondary education through verification interviews.

Federal regulators are investigating how Chinese companies circumvent export restrictions through overseas smuggling and remote compute rental.

New constitutional approach cuts biology-related refusals by 85%, showing how safety classifiers perform better with careful retraining.

Advocates argue a sandbox escape involving advanced models proves voluntary safety measures are insufficient for frontier AI systems.