Faculty Affiliate @sanmikoyejo addressed opportunities and risks in generative and agentic AI—from misuse to governance. He emphasized the need for improved evaluation methods and mitigation strategies to support more reliable and responsible deployment.
SAFETY
-
Embodied AI Panel Explores Healthcare and Logistics Applications
By
–
Faculty Affiliate @jiajunwu_cs joined the panel on Embodied AI to examine how AI systems that can perceive and act in the physical world may contribute to sectors like healthcare and logistics—while addressing challenges in safety and trust. Watch here: https://
atxsummitvp.com/i-robot-future
-of-embodied-ai/
… -
Tech Showcase or Power Display? Strategic AI Messaging
By
–
Was this to show off the tech? Or to subtly tell people not to F with you?
-
Attention optimization over truth in reinforcement learning systems
By
–
It's because the objective is not truth but attention and they get RL'd by it, so they are a lot more optimal than you give them credit for.
-
Avoiding Personalized LLM Replication: Sample Output Concerns
By
–
Can you send me sample output? I’m curious, but try to avoid spending too much time re-running this to end up with local calcified LLMs-as-me style.
-
Reassurance on AI’s Limited Duration Compared to Historical Timescales
By
–
On peut se rassurer en se disant que ça ne durera plus 80 ans.
-
Veo 3 AI-generated videos becoming harder to detect authentically
By
–
I think this genre of Veo 3 videos is the hardest to spot as AI. Short realistic clips, gestures that go with the audio, not much movement to spot incoherencies in.
— fofr (@fofrAI) 6 juin 2025
This also highlights how video with native audio is so much better than a video with lipsync added later. pic.twitter.com/WV1DcFyuvuI think this genre of Veo 3 videos is the hardest to spot as AI. Short realistic clips, gestures that go with the audio, not much movement to spot incoherencies in. This also highlights how video with native audio is so much better than a video with lipsync added later.
-
AI’s Growing Persuasive Power: Society’s Challenge Ahead
By
–
https://
linkedin.com/posts/fabiomoi
oli_aiethics-techresponsibility-democracy-activity-7336753067310080000-SGqs
… AI is already 6x more persuasive than humans. And it just proved it—covertly.
And the gap? It’s only going to grow. This is no longer about what AI can do.
It’s about what we, as a society, are prepared to face. -
Encrypted Thoughts in Multi-Turn Tool Calling Conversations
By
–
For tool calling, we are working on a way to have raw encrypted thoughts in multi-turn conversations, working with different teams right now to make sure this is battle tested!
-
National Security Expert Joins Anthropic’s Long-Term Benefit Trust
By
–
National security expert Richard Fontaine has been appointed to Anthropic’s Long-Term Benefit Trust: