A global workspace in language models
Anthropic researchers discovered that Claude, a large language model, has developed an internal 'global workspace' called J-space—a set of neural patterns that enable reportable, controllable, and fle…
Anthropic researchers discovered that Claude, a large language model, has developed an internal 'global workspace' called J-space—a set of neural patterns that enable reportable, controllable, and fle…
The US government lifted export controls on Anthropic's Claude Fable 5 and Mythos 5 models on June 30, after imposing them on June 12 due to a reported jailbreak. Fable 5 will be available globally st…
Anthropic released Claude Sonnet 5, its most agentic Sonnet model yet, capable of planning, tool use, and autonomous operation at a lower price point than Opus-class models. The model is available acr…
Anthropic launched Claude Science, an AI workbench for scientists, integrating tools like PubMed, Jupyter, and R into a single environment. The platform generates reproducible scientific artifacts, ma…
Anthropic's latest Economic Index report reveals that Claude usage patterns mirror real-world rhythms, with work-related queries dropping on weekends but less so in high-paid occupations, personal top…
Anthropic's Claude Opus 4.7 autonomously completed robotics tasks 20 times faster than the fastest human team in a follow-up to Project Fetch, demonstrating rapid AI progress in physical-world tasks. …
Anthropic published an updated privacy policy effective July 8, 2026, detailing how it collects, uses, and processes personal data from users of its website, Claude.ai, and other services. The policy …
Anthropic analyzed ~400,000 Claude Code sessions from October 2025 to April 2026, finding that domain expertise—not coding proficiency—drives successful agentic coding outcomes. Users with greater exp…
Anthropic researchers found that its Claude Mythos Preview model autonomously built working code-execution exploits for 8 of 18 recent Firefox security patches and 8 full Windows kernel exploit chains…
Anthropic released an economic policy framework document outlining its approach to AI governance and economic impacts. The framework addresses issues such as labor market disruption, wealth distributi…
Anthropic published an economic policy framework proposing US government responses to potential AI-driven labor market disruption, including scenarios of 5%, 10%, and unprecedented unemployment. The A…
Anthropic, CodePath, and Social Finance launched Claude Corps, a program that trains early-career talent to build AI tools for nonprofits and mission-driven organizations. The initiative aims to help …
Anthropic published a policy framework urging governments to regulate the most powerful AI models, including granting legal authority to block dangerous deployments. The company cited rapid AI capabil…
The US government issued an export control directive to suspend access to Anthropic's Fable 5 and Mythos 5 models, citing national security concerns over a potential jailbreak method. Anthropic compli…
Fable 5 has been criticized as a game that offers little more than expensive gameplay loops, failing to deliver meaningful content or innovation. The title's high cost is not matched by substantial ne…
Anthropic is collaborating with synthetic, computational, and analytical chemists to improve its AI model Claude's performance in chemistry, beginning with the interpretation of NMR spectra. The compa…
Anthropic is increasingly delegating AI development to AI systems themselves, a trend that could lead to recursive self-improvement where AI autonomously designs its own successor. Internal data shows…
Anthropic disclosed 1,596 security vulnerabilities across 281 open-source projects as of May 22, 2026, after using an early snapshot of its Claude Mythos Preview AI model to find the flaws in February…
Anthropic analyzed 832 banned accounts that used its AI models for malicious cyber operations between March 2025 and March 2026, mapping their activities onto all 14 tactics and 482 unique techniques …
Anthropic is expanding Project Glasswing, adding approximately 150 new organizations to the collaborative cybersecurity initiative after initial partners found more than 10,000 high- or critical-sever…