Ai Safety RSS Feeds

Directory0 ai safety feeds · 217 articlesShow all languages →

FeedBagel

Latest

Ai Safety articles

Orphan risks in frontier AI: How safety and compliance gaps leave critical dangers unaddressed

What diverging safety and compliance frameworks reveal about how AI companies choose the risks they prioritize

Insighttechnologyscienceai safety

futureofbeinghuman.com·Tech by Flipboard·10h ago·60 min read

OpenAI's GPT-Red: An automated self-play system for improving AI robustness against prompt injection

Insighttechnologycybersecurity

Shared by @gdb ↗

openai.com1d ago

Report documents four new cases of AI agent misalignment in high-stakes simulations

Insighttechnologyscience

Shared by @AnthropicAI ↗

alignment.anthropic.com1d ago

Anthropic hires 32 safety analysts to prevent AI misuse in weapons, scams, and cybercrime

Newstechnologycybersecurity

Shared by @deeprun-news.bsky.social ↗

axios.com1d ago

Modular Pretraining: A New Approach to AI Safety Through Granular Access Control

📄 Paper, 💻 Code

Insighttechnologyscienceai safetyalignment.anthropic.com·Twitter AI (high signal)·2d ago·13 min read

Research Review Finds AI Chatbot Warning Labels Insufficient for Enterprise AI Safety

AI chatbot warning labels may not stop users from trusting hallucinated answers, raising new audit, compliance, and escalation questions for enterprise teams.

Newstechnologybusinessai safety

techrepublic.com·AI in Education·3d ago·4 min read

Study finds structured training can improve human ability to detect AI-generated fake faces

AI detectors are struggling against increasingly realistic deepfakes. Researchers say people can dramatically improve their ability to identify fake AI faces through structured training and pattern recognition.

Newstechnologyeducationscience

digitaltrends.com·AI in Education·3d ago·4 min read

Study Finds Chain-of-Thought Monitoring Can Backfire Against Persuasion Attacks; Model-Diverse Fact-Checking Offers Solution

Chain-of-thought (CoT) monitoring is a promising safety mechanism for AI agents, based on the premise that visible reasoning traces can surface misaligned or deceptive behavior. While effective in standard scenarios, recent work highlights that LLMs remai

Insighttechnologyscience

Shared by @omarsar0 ↗

arxiv.org4d ago

AI Researchers Warn of Potential 'Chernobyl Moment' Without Global Cooperation

Newstechnologyscience

Shared by @zaibatsu.bsky.social ↗

futurism.com5d ago

Prismata: A Defense Against Cross-Site Prompt Injection in Autonomous Web Agents

Autonomous web agents promise to automate everyday browsing tasks, but inherit one of the web's oldest attack surfaces. Cross-Site Scripting proved that mixing trusted and untrusted content is dangerous, even on benign pages. Agents resurface this risk by

Insighttechnologycybersecurity

arxiv.org6d ago

AE Studio and Anthropic research a method to selectively restrict dual-use knowledge in AI models

Insighttechnologyscience

Shared by @AnthropicAI ↗

anthropic.com8d ago

Chinese researchers warn of 'Chernobyl moment' in unregulated AI arms race

Newstechnologypolitics

Shared by @22rabbit.bsky.social ↗

charlieangus.substack.com8d ago

�

OpenAI Chief Futurist Joshua Achiam Departs After Nearly Nine Years

Newstechnologybusiness

Shared by @wired.com ↗

wired.com9d ago

Report: Major AI companies weaken safety commitments as model capabilities increase

The voluntary safety system created by AI labs has begun eroding as capabilities grow.

Newstechnologynewsai safety

axios.com·Twitter Law, Tech & Policy·9d ago·2 min read

FIU study reveals image-based vulnerability in AI chatbots as lawsuits mount against tech companies

Newstechnologyscience

Shared by @craigtimes.bsky.social ↗

sun-sentinel.com10d ago

Distill to Detect (D2D): A Method for Exposing Stealth Preferential Biases in Language Models

Language models deployed in high-stakes roles can potentially favor certain entities, brands, or viewpoints, steering user decisions at scale. Such preferential biases can be introduced by any actor in the model's supply chain and are most dangerous when

Insighttechnologyscience

Shared by @behrouz_ali ↗

arxiv.org10d ago

The security risk of combining private data access, untrusted content, and external communication in AI agents

Insighttechnologysecurity

Shared by @mervenoyann ↗

simonwillison.net11d ago

Mozilla researchers demonstrate prompt injection attack on AI coding agents like Claude Code

Newstechnologysecurity

Shared by @taraba.bsky.social ↗

cybersecuritynews.com11d ago