Risk-management is a topic tracked in our intelligence system with 5 linked articles.
OpenAI's Head of Safety is leaving; signals organizational shift as safety and research teams are being integrated.
Anthropic released Claude Fable 5, the first Mythos-class model widely available, touting top performance and enabled by new safeguards that block high-risk outputs.
A US-sanctioned currency exchange claims a $15 million cyber heist was carried out by unfriendly states, with Grinex saying hacking resources are restricted to such states.
Court filings allege Meta downplayed child harms and hid safety issues, backed by internal docs and testimony in a sprawling multidistrict lawsuit against multiple social platforms.
Adversarial poetry can jailbreak LLMs with high success across many models, revealing significant safety/alignment vulnerabilities and cross-domain risk exposure.
A court ruled Google is liable for AI-generated false statements, signaling that AI designers/operators can bear liability for their outputs.
Subscribe for real-time topic updates and unlimited access to our intelligence platform.