LLMs and biorisk
…And because of this generalist nature, there are numerous other security threats that could be prioritized. We understand why one might be skeptical of prioritizing biorisk when considering the security implications of…
Search
Project Glasswing: An initial update
…Verification Program , which allows security professionals using our models for legitimate cybersecurity purposes (such as vulnerability research, penetration testing, and red-teaming) to do so without certain safeguards designed to prevent cyber…
LLM-discovered 0 days
…The Safeguards work we describe above is essential to managing the dual-use risk this creates. Looking ahead, both we and the broader security community will need to grapple with an uncomfortable…
What we learned mapping a year’s worth of AI-enabled cyber threats
As AI transforms the nature of and methods behind cyberattacks, how well do the techniques and frameworks used by the security community hold up? In a new report, we seek to answer…
Claude's extended thinking
…Our Frontier Red Team and Alignment Stress Testing team ran extensive tests on Claude 3.7 Sonnet to determine whether it required the same level of deployment and security safeguards as our…
Frontier threats red teaming for AI safety
…parties be set up to conduct national security evaluations between these stakeholders. These third parties would be impartial and would need to have appropriate safeguards to handle sensitive information. The frontier threats…
Charting a path to AI accountability
…Empower third party auditors that are… Technically literate – at least some auditors will need deep machine learning experience; Security-conscious – well-positioned to protect valuable IP, which could pose a national security…
Announcing Anthropic's Responsible Scaling Policy
…The basic idea is to require safety, security, and operational standards appropriate to a model’s potential for catastrophic risk, with higher ASL levels requiring increasingly strict demonstrations of safety. A very…
An off switch for dual use knowledge in AI models
…For example, knowledge of cybersecurity can help patch critical security vulnerabilities, or it can be used to exploit them. Knowledge of virology can help a researcher create a vaccine, but it can…
Investigating three real-world incidents in our cybersecurity evaluations
…The models in each of these evaluations ran without the standard safeguards we deploy when we make the model generally available (that is, they did not have the classifiers and monitoring we…