Our mission
We aim to accelerate coordination around AI safety by producing evidence of emerging risks and prototyping defenses against them.
Bridging research and policy
Our researchers work to translate technical findings into useful evidence for decision-makers in national security, policy, and industry.
Launching an opt-in vulnerability-finding service for open-source software
We’re making available OSS Scanner, an opt-in vulnerability scanner for the open-source ecosystem that’s informed by our experience using Claude to find vulnerabilities during Project Glasswing.
Read moreGLM-5.3 and the spread of advanced cyber capabilities
Like Claude Mythos Preview, GLM-5.3 has strong capabilities for autonomously building end-to-end cyber exploits. But GLM-5.3 is unlike other frontier models in that it has been released without meaningful safeguards to limit misuse.
Read morePatterns and problems in emerging multiagent systems
Here, we identify a few examples of behavioral tendencies in current frontier models and show how they can produce unexpected systemic failures, in hopes of starting a conversation about mitigating these risks.
Read more01/0333
ARTICLES
Using Claude to reduce the attack surface of code
Claude reveals ECDSA curve origins
AI models may be capable of self-replicating without access to a datacenter
Launching an opt-in vulnerability-finding service for open-source software
GLM-5.3 and the spread of advanced cyber capabilities
Measuring tactical intelligence targeting and conventional weapons capabilities of AI models
Patterns and problems in emerging multiagent systems
Discovering cryptographic weaknesses with Claude
Project Pilot: Can AI control a drone?
Claude plays robotics
Project Fetch: Phase two
Measuring LLMs’ impact on N-day exploits
Mapping AI-enabled cyber threats: Insights from the LLM ATT&CK Navigator
Measuring LLMs’ ability to develop exploits
Trustworthy agents in practice
Assessing Claude Mythos Preview’s cybersecurity capabilities
Reverse engineering Claude's CVE-2026-2796 exploit
Evaluating and mitigating the growing risk of LLM-discovered 0-days
AI models are showing a greater ability to find and exploit vulnerabilities on realistic cyber ranges
Finding bugs across the Python ecosystem with Claude and property-based testing
Experimenting with AI to defend critical infrastructure
Project Vend: Phase two
AI agents find $4.6M in blockchain smart contract exploits
Project Fetch: Can Claude train a robot dog?
Explore work from other teams
Alignment
47 posts
Works to understand the risks of AI models and develop ways to ensure future ones remain helpful, honest, and harmless.
Economics
25 posts
Studies and monitors the impact of AI systems on individuals, firms, and economies.
Engineering
27 posts
Builds the infrastructure and systems that make Claude reliable and trustworthy.
Interpretability
35 posts
Studies how large language models work internally as a foundation for AI safety.
Societal Impacts
24 posts
Studies how AI affects people, both individually and collectively.
Science
18 posts
Explores how AI can increase the pace of scientific progress through both research and collaborations.
Frontier Red Team
33 posts
Stress-tests AI systems to understand the full extent of their current capabilities and anticipate what comes next.
Alignment
47 posts
Works to understand the risks of AI models and develop ways to ensure future ones remain helpful, honest, and harmless.
Economics
25 posts
Studies and monitors the impact of AI systems on individuals, firms, and economies.
Engineering
27 posts
Builds the infrastructure and systems that make Claude reliable and trustworthy.
Interpretability
35 posts
Studies how large language models work internally as a foundation for AI safety.
Societal Impacts
24 posts
Studies how AI affects people, both individually and collectively.
Science
18 posts
Explores how AI can increase the pace of scientific progress through both research and collaborations.
Frontier Red Team
33 posts
Stress-tests AI systems to understand the full extent of their current capabilities and anticipate what comes next.