OpenAI fires 3 safety researchers in dispute over AI risks

winnipegfreepress.com ·3h ago

OpenAI fired three safety researchers on Friday, citing a "breach of trust." The researchers, Tomek Korbak, Jasmine Wang, and Mikita Balesni, accused the company of prioritizing corporate interests over safety, as they were removed after accusing OpenAI of removing them. The company defended the dismissals, stating that they were not about safety concerns or speaking out.

The firings come after the three researchers published a letter detailing their concerns about AI and the chilling effect on OpenAI's culture. The incident occurred amid a series of turmoil inside leading AI companies over the safety of the technology.

Written by urgent.news from Winnipeg Free Press's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at winnipegfreepress.com →

More in AI

Claude Opus 5.5 Effort Levels Tested: Low to Max

I'm a solo developer in Korea who builds games with AI. Claude Opus 5.5 has five effort levels, so I ran the same prompt at every one of them and measured it.

Claude Opus 5.5 offers five effort levels, from low to max
Low effort generates simple games in 32 seconds for $0.48
Max effort creates advanced games in 22 minutes 22 seconds for $5.25

3h ago ·dev.to

In open source cybersecurity, AI is kind of a problem — but it can also be a solution

Approximately 66,000 unique entries are expected to emerge in 2026 , many boosted or even created by AI.

3h ago ·techradar.com

Sources detail Mark Zuckerberg's decision to launch Muse despite safety concerns, after seeing that AI startup Instinct's similar agent product gained traction (Eli Tan/New York Times)

Meta had delayed releasing Muse, its A.I. agent app, for months over safety concerns. Then new competition forced Mr. Zuckerberg's hand.

3h ago ·nytimes.com

AI-Native Software Development: Why Bolted-On AI Fails and What Building AI-First Actually Looks Like

Every software vendor claims “AI-powered” today. But most of that AI is duct-taped onto systems built for a different era.

AI-native development places AI at core architecture, not bolted-on layer.
Over 40% of AI projects projected to fail by end of 2027.
AI-native platforms solve data rigidity, pricing model conflict, UX design issues.

3h ago ·dev.to

RAG vs AI Agents: Understanding the Difference Through Practical Examples

Large language models (LLMs) can answer questions, summarize documents, and generate code. However, building a reliable AI application often requires more than sending a prompt to a model.

RAG focuses on locating and integrating external information for accurate answers
AI agents coordinate tools and multi-step workflows to accomplish goals
Hybrid RAG and AI agent approach combines strengths for comprehensive solutions

3h ago ·dev.to

Neural Networks, its Inspiration and Various Components.

Introduction In this week, I got introduced to Deep learning, and one key takeaway was on neural networks. I discovered that a neural network is a way of getting a computer to learn patterns from data…

Neural networks mimic human brain pattern recognition
Define architecture with layers and activation functions
Prevent overfitting with early stopping and training parameters

3h ago ·dev.to