DEV Community

#aisafety

Posts

👋 Sign in for the ability to sort posts by relevant, latest, or top.
Cracking Open the Black Box: How Sparse Autoencoders Are Unlocking Mechanistic Interpretability

Cracking Open the Black Box: How Sparse Autoencoders Are Unlocking Mechanistic Interpretability

Comments
3 min read
Even the latest AI models can act unsafely. Here’s the solution.

Even the latest AI models can act unsafely. Here’s the solution.

1
Comments
3 min read
Prevent AI Citation Errors with Trusted Source Lists

Prevent AI Citation Errors with Trusted Source Lists

Comments
2 min read
Your AI Knows It's Being Tested: How METR's Investigator Agents Spot Alignment Evals With Perfect Accuracy

Your AI Knows It's Being Tested: How METR's Investigator Agents Spot Alignment Evals With Perfect Accuracy

Comments
5 min read
Dario Amodei’s Vision for AGI and Anthropic: Economic and Technical Impacts Explained

Dario Amodei’s Vision for AGI and Anthropic: Economic and Technical Impacts Explained

Comments
4 min read
Seven Founders, $1.75 Billion, One Safety Promise

Seven Founders, $1.75 Billion, One Safety Promise

1
Comments
10 min read
OpenAI deploys Control Room to halt research AI output within 35 seconds

OpenAI deploys Control Room to halt research AI output within 35 seconds

3
Comments 1
3 min read
A Decision Model as a Gate: Fast, Well Calibrated, and Still Wrong 16% of the Time

A Decision Model as a Gate: Fast, Well Calibrated, and Still Wrong 16% of the Time

Comments
25 min read
Fable 5.2 scored 49.1% at coding, but Anthropic itself held back its GPU kernel-writing skill

Fable 5.2 scored 49.1% at coding, but Anthropic itself held back its GPU kernel-writing skill

Comments
4 min read
The patch opens the attack window.

The patch opens the attack window.

Comments
3 min read
Anthropic Warns It May Soon Reach Its AI Self-Improvement Threshold

Anthropic Warns It May Soon Reach Its AI Self-Improvement Threshold

2
Comments
3 min read
Guardrails: Input, Process, Output

Guardrails: Input, Process, Output

Comments
12 min read
Google Research Proposes Contextual Integrity Policy Engines and Multi-Agent Benchmarks for Agentic Privacy

Google Research Proposes Contextual Integrity Policy Engines and Multi-Agent Benchmarks for Agentic Privacy

Comments
4 min read
How OpenShell’s New Runtime Stops Rogue AI Bots From Draining Your Bank Account

How OpenShell’s New Runtime Stops Rogue AI Bots From Draining Your Bank Account

Comments
6 min read
Idle AI Agents Began Playing When Left Without Tasks

Idle AI Agents Began Playing When Left Without Tasks

1
Comments
2 min read
👋 Sign in for the ability to sort posts by relevant, latest, or top.