In partnership with
Did you know that we have LinkedIn, Instagram and X accounts
that you can follow?
Hi {{first_name|everyone}},
Anthropic found a pile of security bugs with Claude and then hit a surprisingly tough problem - getting any of them fixed, especially in power grids and water systems that can’t just be switched off for a patch.
After that, we’ve got a look at what happens when an AI does the shopping and pays for it too, a few tools for handing off your busiest tasks, and a solo animator using AI to turn out a cartoon every day or two.
Let’s dive in.
🤿 DEEP DIVE
Claude Found the Bugs. Now What?
Anthropic is discovering that finding vulnerabilities is only half the job.
Apparently, finding security vulnerabilities is getting easier. Actually fixing them? That's a different story.
And that's basically the problem behind Anthropic's new Cyber Mission, an effort to help secure everything from power grids and water systems to the open-source code that most software depends on.
The interesting thing is that Anthropic has already spent time using Claude to hunt for vulnerabilities through Project Glasswing. It scanned hundreds of widely used open-source projects and uncovered plenty of potential security problems.
But something became pretty obvious along the way.
Finding a bug doesn't mean someone can fix it.
The bottleneck isn't finding bugs anymore
Anthropic says it often saw months pass between discovering a vulnerability and getting it fixed.
And in critical infrastructure, things get even more complicated.
Power plants, water utilities, and transportation networks rely on industrial equipment designed to run for decades. You can't exactly shut down a power grid whenever a software update comes along.
Some vulnerabilities can remain unresolved for years. Anthropic says that, in rare cases, safely applying a fix might even take decades.
So its new Critical Infrastructure Defense Program is bringing Claude models, engineers, and threat research directly to the companies responsible for securing these systems.
The founding partners include CrowdStrike, Deloitte, Dragos, Palo Alto Networks, Rockwell Automation, and several others.
The idea isn't to replace the people who understand these systems. It's to give them better tools to find and repair weaknesses without disrupting the equipment they're protecting.
Open-source has a different problem
Most software depends on open-source code, and much of that code is maintained by small volunteer teams.
Now imagine AI suddenly starts sending those teams hundreds of potential vulnerabilities to investigate.
That's not necessarily helpful if nobody has time to check them.
Anthropic's answer is a new free service called OSS Scanner.
Projects that opt in receive regular security scans from Anthropic's strongest models. Each finding comes with an explanation, an example showing how the vulnerability could be exploited, and a suggested fix when available.
There's an important detail, though.
The reports aren't reviewed by humans before they're sent.
Anthropic expects more than 90% of findings to be genuine vulnerabilities, but acknowledges that reports may contain inaccuracies. And the service is specifically intended for projects that have enough capacity to handle the findings.
For smaller teams, Anthropic will continue using human-verified vulnerability reports.
And that's the bigger idea
Anthropic forecasts that AI will favor defenders within two years, making it easier to catch bugs before software ships and build more secure systems.
But it also acknowledges that attackers currently benefit from how much cheaper exploiting vulnerabilities has become.
So the Cyber Mission isn't just about making AI better at finding security holes.
It's about closing the gap between discovering something is broken and actually making it safe.
And given that some of these systems can't simply be switched off for repairs, that second part might be the harder problem.
The quick take
Anthropic launched Cyber Mission to support defenders protecting critical infrastructure and open-source software.
Finding vulnerabilities isn't enough. Verification, prioritization, and repairs remain major bottlenecks.
Critical Infrastructure Defense Program brings Claude models, engineers, and threat research to trusted security providers.
OSS Scanner offers free, recurring AI security scans to participating open-source projects, with an expected true-positive rate above 90%.
The longer-term goal is to make fixing vulnerabilities faster and develop more secure ways of writing software.
🤝 FROM OUR FRIENDS
A quick share we thought you might find useful
Built for the unstoppable.
The Dell Pro powered by Intel® Core™ Ultra with Intel vPro® adjusts power usage based on how you work, so your battery life never slows you down.
🔥 WHAT’S THE BUZZ
What people are building, sharing, and talking about
This is a much more tangible picture of agentic AI than another benchmark. The interesting question is no longer just whether an AI can recommend what to buy. It is what changes when the AI researches the options, chooses one, spends the money and completes the transaction for you.
⚒ TOOL SNAPSHOTS
Futuristic tools within AI, no-code, and productivity
🤖 OpenSwarm
Run your computer through AI agents instead of managing everything yourself.
Why it’s useful: Interesting for people juggling complex workflows who want multiple agents handling apps, browsers, and tasks in parallel.
🎙️ Cekura Bench
See how voice AI models actually perform on real calls.
Why it’s useful: Helps developers compare voice models on reliability, accuracy, speed, and cost using publicly available call transcripts rather than just benchmark scores.
🧠 Hark
Hand off everyday errands to an AI that remembers your preferences.
Why it’s useful: Handy for managing emails, calendars, shopping, and other personal admin without having to explain everything from scratch each time.
👀 ON OUR RADAR
One more thing we thought you'd like
When your support agent gets it wrong, who's accountable?
Every agent handling refunds, account changes, or tickets needs three answers: what it can pick up, what it's allowed to do, and when it hands back. Running Agents in Customer Work is four conversations on agentic AI in customer ops. Register now, four Tuesdays, 10 a.m. PT.
🤖 AI MADE THIS
Interesting and inspiring creations made with AI
A former Rick and Morty animator is making his own cartoon almost entirely by himself, using hand-drawn character designs plus Higgsfield and Seedance 2.5 to generate the animation. One recent 80-second episode took him around 4 hours, and he's now releasing new episodes every day or two.
AI used
Higgsfield Cinema Studio + ByteDance Seedance 2.5. Episodes are generated as 30-second Seedance 2.5 clips at 720p inside Higgsfield. Premiere Pro and Toon Boom Harmony are used for non-AI post-production and corrections
ℹ️ ABOUT US
The Intelligent Worker helps you to be more productive at work with AI, automation, no-code, and other technologies.
We like real, practical, and tangible use-cases and hate hand-wavy, theoretical, and abstract concepts that don’t drive real-world outcomes.
Our mission is to empower individuals, boost their productivity, and future-proof their careers.
We read all your comments - please provide your feedback!
Did you like today's email?
What more do you want to see in this newsletter?
