TLDRocket
Sign in

Safety & Ethics

430 summarised stories in Safety & Ethics, each linking back to the original source. Browse all topics →

Wednesday, 19 November 2025

Strengthening our safety ecosystem with external testing

OpenAI Blog 8 months ago

OpenAI has expanded its practice of engaging independent experts to evaluate its advanced AI systems for safety and risk assessment. The company conducts third-party testing to validate its existing safeguards and improve transparency around how it measures model capabilities and potential harms. This external evaluation approach allows OpenAI to identify vulnerabilities and strengthen its overall safety protocols before deployment.

GPT-5.1-Codex-Max System Card

OpenAI Blog 8 months ago 3 sources

OpenAI released a system card for GPT-5.1-Codex-Max that documents safety measures across model and product layers. The safeguards include specialized safety training to prevent harmful task execution and prompt injection attacks, along with agent sandboxing and configurable network access controls. These mitigations aim to reduce risks when the system handles sensitive code generation and autonomous agent operations.

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads 60+ sources, removes duplicate coverage, and summarises the day in two minutes. Free, no spam, unsubscribe anytime.