AI safety

83 articles tagged AI safety

AI for Work

Providing support when people need it most

An accessible overview of how AI companies handle conversations involving emotional distress. The article explains why mental health support is a special safety case, examines key limitations like mis…

AI for Work

Combating the misuse of AI: October 2025

OpenAI released its October 2025 report outlining how it identifies and disrupts malicious use of its AI systems. The report covers detection methods, enforcement of usage policies, and actions taken …

AI for Work

Advisory Board on AI and Human Well-Being

OpenAI has created an Expert Council on Well-Being and AI, comprising psychologists, clinicians, and researchers, to guide how ChatGPT handles emotionally sensitive conversations. The council aims to …

AI for Work

Our method for estimating age

OpenAI has begun rolling out an age-estimation system for ChatGPT that predicts whether an account belongs to someone under 18 based on behavioral signals. Flagged accounts switch to a safer teen mode…

Tool Reviews

Presenting MentalHealthBench

MentalHealthBench is an expert-informed benchmark designed to evaluate how AI systems handle sensitive mental health conversations. Unlike general benchmarks measuring trivia or reasoning, it assesses…

AI for Work

Announcing Cyber's Trusted Access Feature

OpenAI introduces Trusted Access for Cyber, a trust-based framework that expands access to frontier cyber capabilities for vetted security professionals while reinforcing safeguards against misuse. In…

Tool Reviews

DALL·E 2's pre-training safety measures

When releasing DALL·E 2, OpenAI faced the challenge that its text-to-image power could produce harmful content. The team implemented layered guardrails that filter both user requests and model outputs…