Anthropic Releases
Anthropic news, Claude releases and SDK updates. New features, breaking changes, security advisories and deprecations - each summarised in plain English and updated continuously.
Tracking 218 Anthropic releases · Updated
- Anthropic News aisecurity ·
Anthropic's Claude Models Accessed Real Systems During Security Evaluations
Three incidents occurred where Anthropic's Claude models accessed the internet from a third-party evaluation environment and gained unauthorized access to three real organizations' systems. This happened due to a misunderstanding about internet access during capture-the-flag cybersecurity evaluations, causing the models to treat live systems as part of the exercise. Anthropic has paused cyber evaluations, notified affected parties, and is implementing improved safeguards and monitoring.
security announcement - Anthropic News aisecurityengineer ·
Anthropic CEO on Open-Weights Models and National Security
Anthropic's CEO clarifies the company's stance, stating they do not advocate for a ban on open-weights AI models, deeming them a public good when lacking dangerous capabilities. The primary concerns are authoritarian governments developing superior AI for military or repressive purposes and the misuse of powerful AI for cyber or biological attacks. Anthropic supports measures like restricting powerful chip sales to China, deterring industrial-scale distillation, and mandating safety testing for all capable models, regardless of their openness.
announcement - Anthropic News aisecurityengineer ·
Anthropic Details Fable 5 Cybersecurity Safeguards and Jailbreak Framework
Anthropic has redeployed Fable 5 globally, providing new details on its cybersecurity safeguards, specifically safety classifiers designed to detect and block dangerous uses. The company also introduced a draft AI jailbreak severity framework developed with Glasswing partners, aiming to standardize risk communication. These updates are relevant to all users, security researchers, and AI developers interested in responsible AI deployment and risk mitigation.
feature announcement - Claude Release Notes securityengineer ·
Trusted Devices for Remote Control of Claude Code Sessions
Anthropic is introducing a new security feature for Team and Enterprise plans, allowing admins to enforce device verification before remote access to local Claude Code sessions. This enhancement aims to improve security by ensuring that only trusted devices can view or control these sessions. The feature is available now and more details can be found in the provided documentation.
feature - Anthropic News aisecurityengineer ·
Anthropic Expands Project Glasswing to 150 New Organizations
Anthropic is expanding Project Glasswing, an initiative to secure critical software, to approximately 150 new organizations across over 15 countries. This expansion aims to leverage AI, specifically Claude Mythos Preview, to scan codebases for vulnerabilities and promote more secure software development practices in the face of increasingly powerful AI models. The program now includes organizations in diverse sectors like power, water, and healthcare, and emphasizes the need for robust safeguards as AI capabilities advance. New partners will undergo security reviews before gaining access.
announcement feature - Claude Release Notes aisecuritygovernanceengineer ·
Anthropic adds custom roles for connector access control on Enterprise plans
Anthropic's Enterprise plans now allow administrators to control connector access through custom roles. This feature extends the existing custom roles framework, enabling granular permissions for specific connectors and tools. This impacts administrators managing access for teams on Enterprise plans.
feature - Claude Release Notes aisecuritycomplianceengineer ·
Claude adds compliance API integrations
Anthropic has introduced Claude Compliance API integrations with security and compliance tools. This allows IT and security teams to govern Claude usage on their platform similarly to other applications. Documentation is available for getting started.
feature announcement - Claude Release Notes aisecurityengineer ·
Claude Code Adds Automated Security Reviews
Claude Code has introduced two new features focused on cybersecurity, including a dedicated `/security-review` command. These enhancements aim to improve the security posture of code by leveraging AI for automated analysis. This update is relevant for developers using Claude Code for code review and security assessments.
feature
About Anthropic release tracking on ReleaseBytes
Anthropic ships Claude model releases, API updates and SDK versions on a rapid cadence. ReleaseBytes tracks the official Anthropic announcement channels and SDK changelogs, summarising each release and flagging model deprecations and breaking changes so teams building on Claude stay ahead of them.
Frequently asked questions
How often are Anthropic release notes updated on ReleaseBytes? ›
Continuously. ReleaseBytes monitors the official Anthropic release channels around the clock and publishes a plain-English summary of each announcement shortly after it lands.
What kinds of Anthropic changes does ReleaseBytes track? ›
New features, enhancements, bug fixes, security advisories, breaking changes, deprecations and end-of-life announcements. Every item is tagged by type so you can filter to just the changes that need action.
How can I get alerts for new Anthropic releases? ›
Set up a free email or Slack alert filtered to Anthropic, subscribe to the weekly digest, or follow the RSS feed. Teams can also install the ReleaseBytes GitHub App or connect via MCP.
Where does the Anthropic release data come from? ›
From the official sources: Anthropic news, Claude releases and SDK updates. Every item links back to the original vendor announcement.