We’re opening applications for the next two rounds of the Anthropic Fellows Program, beginning in May and July 2026. We provide funding, compute, and direct mentorship to researchers and engineers to work on real safety and security projects for four months.
SAFETY
-
GWM-1 Simulates Human Behavior in Immersive Environments
By
–
GWM-1 can simulate environments, physics and motion. But one of the most complex things to simulate is human behavior – the way people look, move and respond in conversation in a way that feels natural and immersive. This is really difficult to get right.
— Runway (@runwayml) 11 décembre 2025
We believe we’ve… pic.twitter.com/QyDNcH2xiKGWM-1 can simulate environments, physics and motion. But one of the most complex things to simulate is human behavior – the way people look, move and respond in conversation in a way that feels natural and immersive. This is really difficult to get right. We believe we’ve
-
New AI Safety Partnership Focuses on Foundational Research
By
–
We’re also announcing a new partnership with @AISecurityInst that builds on two years working together & will focus on foundational safety and security research essential for realising AI’s potential to benefit humanity.
-
AI Impact on Artists: Ethics and Moral Clarity Debate
By
–
Watching Julian get destroyed by the Machiavellians for making a questionable point badly is bittersweet. He could have made same point about AI & artists (no impact on cancer cures) and been morally justified, but he didn't have the clarity nor conviction at the time / on this.
-
OpenAI hints at ‘garlic’ model ahead of GPT-5.2
By
–
BREAKING : OpenAI hints at “garlic”, a rumoured codename of their upcoming model. GPT-5.2 is expected tomorrow
-
Google upgrades Stitch Agent with Gemini 3 Pro
By
–
Google upgraded Stitch design Agent with Gemini 3 Pro, which is the new default mode now.
— 🚨 AI News | TestingCatalog (@testingcatalog) 10 décembre 2025
Testing time 👀 https://t.co/uW1w4r2fOe pic.twitter.com/oAaZrtvFTeGoogle upgraded Stitch design Agent with Gemini 3 Pro, which is the new default mode now. Testing time
-
Grok AI walks through saving wife’s life with CPR guidance
By
–
Someday your robot might save your life. But this is a huge step. Have you ever asked AI about different crises? I have. “Hey @grok my wife just fell on the floor and isn’t breathing, what should I do?” It walks through saving her life even if I didn’t know CPR. My first
-
Specification-Driven Generation Technique
By
–
Technique 5: Specification-Driven Generation Engineers write a spec first, get model agreement, THEN generate. This separates "what to build" from "how to build it" and catches misalignment early. Template: First, write a specification for [task] including:
– Inputs and their -
Metacognitive Scaffolding for Error Prevention
By
–
Technique 3: Metacognitive Scaffolding Instead of asking for an answer, engineers ask the model to explain its reasoning process BEFORE generating. This catches logical errors at the planning stage. Template: Before you [generate output], first:
1. List 3 assumptions you're -
Multi-Shot Technique with Failure Cases
By
–
Technique 2: Multi-Shot with Failure Cases Everyone uses examples. Engineers show the model what NOT to do. This creates boundaries that few-shot alone can't establish. Template: Task: [what you want] Good example:
[correct output] Bad example:
[incorrect output]
Reason it