
Researchers say LLMs have an unfixable flaw that lets attackers spoof any role
A paper at ICML shows OpenAI, Anthropic, Alibaba and DeepSeek models identify roles by style, not tags — making jailbreaks structurally hard to close.
Tag · 5 stories
Every story tagged GPT-5 on AI Chat Daily.
GPT-5 is OpenAI's current frontier model family, launched in August 2025 and updated through GPT-5.1 in April 2026. We cover every release, pricing move, and capability benchmark — across ChatGPT, the API, and downstream products like Copilot.

A paper at ICML shows OpenAI, Anthropic, Alibaba and DeepSeek models identify roles by style, not tags — making jailbreaks structurally hard to close.

OpenAI says its top reasoning model gave Derya Unutmaz the breakthrough insight that closed a stalled immunology project with cancer implications.

The proof marks the first autonomous AI solution to a prominent open math problem, verified by field experts.

The new voice stack handles 70 input languages and 13 output languages, pushing the API beyond simple call-and-response.
The o-series-style reasoning model lands in ChatGPT with adaptive budgets — and a new opt-in 'extended thinking' toggle.
The tightly edited AI news email engineers, founders, and investors actually open. One email. Every weekday. Five minutes to finish.
The briefing read inside teams at