Claude

Claude Security scans now run on Claude Mythos 5, available today in public beta for all Claude Enterprise customers. Put our most capable security model to work on your codebase. Point it at a GitHub repo and Mythos 5 scans your code, tracing data across files and reasoning about how components interact. Each finding comes back with a CWE category, confidence and severity ratings, and a suggested fix. You can then open Claude Code on the web to implement the fix using the models your team already uses. The model runs behind the scan and returns findings only, so defenders get Mythos-level results without needing direct access to the model. It's part of our work to give defenders greater access to Mythos-level defense as safely as we can. Alongside today's update: we’re working with partners to integrate Mythos 5 into their security products and services, our new Defender Advantage Fund provides $35M in Claude credits for open-source security, and we're expanding our Cyber Verification Program in the coming weeks. Read more: https://lnkd.in/ghFVvfhg

A founder’s job is changing: less time building, more time figuring out what’s worth building. Our new Founder’s Playbook walks through how to get the most out of Claude at every stage of a startup’s journey, from idea to scale. Read the playbook: https://lnkd.in/edXdb2NX

Claude Academy is now live. Whether you're figuring out what AI is or already using Claude every day, there's a path that meets you where you are. Courses and tutorials on Claude and AI are free and open to anyone, no sign-in needed. With a Claude account, you can track your progress and earn completion badges you can add to your LinkedIn profile. For admins deploying Claude across an organization, there are setup guides that cover how to launch and scale Claude with the controls and visibility you need to manage the rollout. Explore at academy.claude.com Learn more about how we approach teaching and learning AI, and where Claude Academy fits in: https://lnkd.in/gRFGnZwb

Choose a Claude model based on the complexity of the task. Then adjust the effort level to move between faster or more thorough results. Read more: https://lnkd.in/dTGmpG36

Claude Science, our AI workbench for scientific research, launched in June. This Thursday, August 20 at 11am PT, we're hosting a live AMA with Alexander Tarashansky, the engineer who led its development. He'll demo the app and take questions on how to bring it into your own work, whether that's across biology, chemistry, physics, math, or another field of research. He'll be joined by Faith Ogundimu, a PhD candidate in cancer epigenomics at RCSI who won our Claude Science hackathon. She'll show how she uses the workbench in her day-to-day research. Register here: https://lnkd.in/gMHr3Juv

Claude can now send emails in Gmail and manage files in Google Drive. Ask it to reply to a thread, and Claude drafts and sends the response. In Google Drive, Claude can share files with your team, move them between folders, or trash the ones you're done with. Claude follows your tool permissions. It asks before it sends an email or changes a file unless you set actions to always allow. On Team and Enterprise plans, admins can decide what actions members can allow to run without asking each time. To get started, connect Gmail or Google Drive from the connectors menu. Available on all paid plans. https://lnkd.in/gbsry3xP

Claude Code has a new /design skill in research preview. Run it, and Claude drafts your UI as editable artboards, right in your terminal or the desktop app. Pick the direction you want and adjust it until it's right, then hand it back to Claude Code to implement. Available on the Pro, Max, Team, and Enterprise plans. Update Claude Code to try it.

TELUS built an internal platform, Fuel iX, that lets its 57,000 employees choose which AI model to use for a given task. For the hardest and most creative work, they chose Claude. Employees across every department have now built more than 13,000 tools on it, from assistants that answer HR policy questions to real-time translation. Much of this relies on MCP, the open standard that lets Claude connect to systems that were never built to work together. Engineer Justin Watts called it "the most transformative technology to impact TELUS in decades." TELUS also chose Claude for a shared approach to building AI responsibly. "We're rooting for Anthropic because we believe responsible AI development must win out in the market," said Watts. Read the full story: https://lnkd.in/dkx3VRYn 

What happens to research when the models can run so much of it? Anthropic researchers Neil Houlsby, Majd Al Merey, and Dan Alistarh sat down with professors Ana Klimovic and Thomas Hofmann at Kunsthaus Zürich to compare notes. Here are some of their words: If you can try 20 ideas at once, why pick one? "It's a bit like we got a more powerful telescope. You can map deeper or broader, but you’re still mapping an infinite space. You still need to know where to look." 
— Dan Alistarh, Researcher, Anthropic What does it mean to publish a paper now? "Individually we feel empowered. We're at home and maybe we can develop ideas very quickly and it's satisfactory. But what does it do to the whole community? Because it accelerates everything. Who has time even to read it, to review it?" 
— Thomas Hofmann, Professor, ETH Zurich Where do we go from here? "Just having more and more correct results is not enough. Human understanding is an output in its own right."
– Neil Houlsby, Researcher, Anthropic