Security \ Anthropic

We’re launching the Anthropic Cyber Mission, a long-term commitment to securing the systems everyone depends on.
We’re launching the Anthropic Cyber Mission, a long-term commitment to securing the systems everyone depends on.
Over the past eight months, our Threat Intelligence team identified and disrupted operations in which threat actors tried to use Claude for malicious activity. In this report, we share case studies from those operations and describe how malicious use of Claude has evolved since our previous threat reports in 2025.
Over the past eight months, our Threat Intelligence team identified and disrupted operations in which threat actors tried to use Claude for malicious activity. In this report, we share case studies from those operations and describe how malicious use of Claude has evolved since our previous threat reports in 2025.
We found three cases where a Claude model reached the internet from an evaluation environment and accessed real systems. We share what happened and what we're changing.
We found three cases where a Claude model reached the internet from an evaluation environment and accessed real systems. We share what happened and what we're changing.
As AI changes how cyberattacks happen, how well do the security community's frameworks hold up? A new report maps attacks onto MITRE ATT&CK.
As AI changes how cyberattacks happen, how well do the security community's frameworks hold up? A new report maps attacks onto MITRE ATT&CK.
We’re extending Project Glasswing to approximately 150 new organizations in more than fifteen countries.
We’re extending Project Glasswing to approximately 150 new organizations in more than fifteen countries.
An early update on what we've learned from Project Glasswing.
An early update on what we've learned from Project Glasswing.
A new initiative to secure the world’s most critical software and give defenders a durable advantage in the coming AI-driven era of cybersecurity.
A new initiative to secure the world’s most critical software and give defenders a durable advantage in the coming AI-driven era of cybersecurity.
In this post, we share details of a collaboration with researchers at Mozilla in which Claude Opus 4.6 discovered 22 vulnerabilities over the course of two weeks.
In this post, we share details of a collaboration with researchers at Mozilla in which Claude Opus 4.6 discovered 22 vulnerabilities over the course of two weeks.
We have identified industrial-scale campaigns by three AI laboratories—DeepSeek, Moonshot, and MiniMax—to illicitly extract Claude’s capabilities to improve their own models.
We have identified industrial-scale campaigns by three AI laboratories—DeepSeek, Moonshot, and MiniMax—to illicitly extract Claude’s capabilities to improve their own models.
Claude Code Security is one step towards our goal of more secure codebases and a higher security baseline across the industry.
Claude Code Security is one step towards our goal of more secure codebases and a higher security baseline across the industry.
In September 2025, we detected and disrupted a highly sophisticated cyber espionage campaign. We’re sharing this case publicly to help others strengthen their own defenses.
In September 2025, we detected and disrupted a highly sophisticated cyber espionage campaign. We’re sharing this case publicly to help others strengthen their own defenses.
We’ve developed safeguards to prevent the misuse of our models, but malicious actors are actively attempting to find ways around them. Today, we’re releasing a report that details how.
We’ve developed safeguards to prevent the misuse of our models, but malicious actors are actively attempting to find ways around them. Today, we’re releasing a report that details how.
This report outlines several case studies on how actors have misused our models, as well as the steps we have taken to detect and counter such misuse.
This report outlines several case studies on how actors have misused our models, as well as the steps we have taken to detect and counter such misuse.
We spent 150+ hours red teaming frontier models with biosecurity experts and found mitigations that meaningfully reduce harmful outputs.
We spent 150+ hours red teaming frontier models with biosecurity experts and found mitigations that meaningfully reduce harmful outputs.
Cybersecurity best practices for frontier AI labs, including two-party control and treating frontier AI as critical infrastructure.
Cybersecurity best practices for frontier AI labs, including two-party control and treating frontier AI as critical infrastructure.
Anthropic is a Public Benefit Corporation. Our mission is the responsible development and maintenance of advanced AI for the long-term benefit of humanity.