Major updates from LlamaCon! We’re advancing AI security with new open-source Llama protection tools and new AI- powered solutions for the defender community. Developers can now access: — Llama Guard 4, a customizable safeguard that supports protections for text and image
SAFETY
-
AI Systems and User Competency Assessment in Decision-Making
By
–
Also note that you will with very high probability interact with processes which explicitly or implicitly have to decide, discretely or continuously, “Is this user competent to make their own decisions?”
-
Dario Amodei Predicts AI Will Become Central Axis of Power
By
–
Il y a quelques jour @DarioAmodei sortait son fameux papier d'appel urgent à tout le secteur de la tech et de l'IA.
— VISION IA (@vision_ia) 29 avril 2025
Dario Amodei ne fait pas des prédictions. Il trace la carte. Ex-DeepMind, ex-OpenAI, fondateur d’Anthropic.
Sa prédiction : l’IA devient l’axe central du pouvoir… https://t.co/oIE80VUc5d pic.twitter.com/3fWW3XJC1cIl y a quelques jour @DarioAmodei sortait son fameux papier d'appel urgent à tout le secteur de la tech et de l'IA. Dario Amodei ne fait pas des prédictions. Il trace la carte. Ex-DeepMind, ex-OpenAI, fondateur d’Anthropic. Sa prédiction : l’IA devient l’axe central du pouvoir
-
Oversight Limits: Can Weaker AI Control Smarter Systems?
By
–
Can a weaker AI really keep a smarter one in check?
New research puts scalable oversight to the test—and the results aren’t all comforting. Even layered oversight struggles once the gap gets too wide. Turns out, keeping super AI honest might be harder than we thought. -
Scaling Laws For Scalable Oversight
By
–
Scaling Laws For Scalable Oversight
Link: https://
arxiv.org/pdf/2504.18530 -
Google Trust Safety Research at CHI2025 Conference
By
–
Come to the #CHI2025 Google booth today during the morning break, where Patrick Kelley will talk about Trust & Safety at Google, including the research we do on topics of child safety, at-risk users, misinformation, and of course, AI safety.
-
Clippy versus Anton: Competing Visions for AI Post-Training Design
By
–
🆕Please stop forcing Clippy on those who want Antonhttps://t.co/efaesB67zo
— Latent.Space (@latentspacepod) 28 avril 2025
ChatGPT-4o's glazing lays open Clippy vs Anton: The two extremes of desires in AI post-training and product pic.twitter.com/6BtR3M67aIPlease stop forcing Clippy on those who want Anton https://
latent.space/p/clippy-v-ant
on
… ChatGPT-4o's glazing lays open Clippy vs Anton: The two extremes of desires in AI post-training and product -
Agentic AI: Enterprise Defense and Risk Mitigation Strategies
By
–
The future of cybersecurity isn’t just AI-powered, it’s agent-empowered. Agentic AI enables new defense capabilities but also new risks. Join us at #RSAC to learn how enterprises can leverage agentic AI and defend against it, reducing risk while scaling security
-
IRB approval questioned for unauthorized Reddit study with bots
By
–
can someone explain to me how something like this — a secret study, unauthorized by reddit, carried out with bots posting comments on a popular subreddit — got IRB approval from the University of Zürich?
-
AI Pioneer Signs Letter Urging OpenAI to Honor Its Mission
By
–
A l'instant : post du "père de l'IA" : J'apprécie la mission d'OpenAI qui est "d'assurer que l'intelligence artificielle générale bénéficie à toute l'humanité", et j'aimerais les empêcher de la vider complètement de sa substance. J'ai signé une nouvelle lettre adressée à