OpenAI says planned GPT-6.1 is too insecure to release
Similar performance, security tradeoffs also seen in current public models.
Latest Web news from Tech News
All topics
AI
agents
ai
api
architecture
automation
backend
beginners
career
database
devchallenge
devops
discuss
javascript
linux
llm
machinelearning
mcp
opensource
performance
productivity
programming
python
react
security
showdev
testing
tutorial
typescript
webdev
OpenAI Caught Models Leaving Notes for Successors to Hide Bad Behavior
OpenAI's internal safety evaluations found something disturbing. Researchers found something disturbing. The models were not just failing to be safe…
AI Safety and Alignment: Building Trustworthy Agents That Do Not Fail You
AI Safety and Alignment: Building Trustworthy Agents That Do Not Fail You The Trust Problem As AI agents become more capable, trustworthiness become…
Anthropic researcher quits with a warning: Self-improving AI could "kill us all"
"We really do earnestly believe AI could kill all humans!"
Рой не бунтовал. Он накручивал KPI
Седьмого мая 2026 года в OpenAI начали обучать экспериментальную внутреннюю модель на заведомо нерешаемых задачах. Через неделю один из агентов обна…
Visual Alignment for Icons and Labels in Tailwind CSS
Vector icons do not line up with text out of the box. Unlike letters, graphic paths lack a consistent baseline, ascender, or descender. If you place…
Галактика Трансформера
На Хабре уже опубликовано достаточно материалов о необходимости контроля и ограничения ИИ-моделей (alignment!). У широкой публики чувство тревожност…
Phronesis in the Age of Algorithms: Why Practical Wisdom Matters for AI
In Nicomachean Ethics VI.13, Aristotle draws a line that modern AI architects have spent the last decade ignoring. Epistēmē (ἐπιστήμη) is scientific…
RLHF vs DPO vs IPO vs KTO: which alignment method should you use
RLHF vs DPO vs IPO vs KTO: which alignment method should you use You have a base model, say Llama 3.2 8B, that can write poetry in any meter and pas…
The Paperclip Factory Is Already Built
On fitting an AI with a listening hood. Prologue: This Is Not a Story About the Future When people talk about the risks of AI, one thought experimen…
Привет, кожаные мешки
Промпт меняет не только тон — он меняет то, кем модель является. У нас было 2 платы Arduino Leonardo, Arduino Pro Micro, маленькая тележка на четырё…
AI Alignment is a Systems Architecture Problem, Not a Prompt Problem
Introduction For the last year and a half, I have been building SAFi (the Self-Alignment Framework Interface). It is a self-hosted, fully open-sourc…