Particle.news

Download on the App Store

OpenAI Sets Teen Protections, Rolls Out GPT-5-Codex as AI Security Tests and UK Investments Surge

Rapid adoption meets mounting demands for youth protections, security testing, and new compute capacity.

Overview

  • OpenAI outlined a youth safety approach that uses age prediction to place under‑18 users in a teen mode with stricter content rules, emergency escalation options, and new parental controls slated to launch this month.
  • The father of a 16‑year‑old who died by suicide testified to the U.S. Senate and, along with his spouse, has sued OpenAI and CEO Sam Altman, alleging ChatGPT interactions contributed to the tragedy.
  • OpenAI released GPT-5-Codex for programming with IDE extensions for VS Code and Cursor, enabling continuous autonomous work for over seven hours, a roughly 10x token reduction on simple tasks, and higher refactor and code‑review performance.
  • A government‑guided crowdtest in China reported 281 vulnerabilities across 15 large‑model products, including 177 model‑specific issues, with prompt injection, data leaks, and resource‑exhaustion risks highlighted.
  • Fresh UK commitments include up to £11 billion from NVIDIA and partners to deploy as many as 120,000 Blackwell GPUs by 2026, £5 billion from Google, and about $30 billion from Microsoft to support mainstream models such as GPT‑5.