Canva launches its own AI image model

Canva wants AI images you can actually edit

Canva just released its first AI image generation model, and it has a practical advantage for people who make designs every day.

It's called Canva Craft.

Unlike typical image generators, Craft is trained on human-made designs, including posters, presentations, menus, and social posts. It can generate images in up to 2K resolution across 30 aspect ratios, with readable text and individual elements that users can edit without changing everything else.

That matters more than it sounds.

I think Canva understands a common frustration with AI images: getting a decent result is easy, but fixing a headline or moving one object can mean starting over.

Craft could make that process easier, especially for people creating marketing materials and presentations on tight deadlines.

There's a business benefit, too. Building its own model could reduce Canva's reliance on external AI providers and give it more control over costs and performance.

The bigger question is whether Canva can make these editable AI designs reliable enough for everyday professional work.

Microsoft's AI is getting better at thinking like a doctor

Microsoft's new medical AI was more than twice as accurate as general physicians at diagnosing complex cases in a controlled study.

Here's what's different.

Instead of receiving all the patient information upfront, the system asks questions, decides which tests to order, and updates its diagnosis as results come in. It was tested on more than 300 unusually difficult medical cases.

I think the interesting part is how the AI decides what information it needs next. In a real consultation, doctors rarely have every test result sitting on their desk, and knowing what to investigate is a major part of the job.

Microsoft also found that its system completed diagnostic work-ups at a lower cost.

That's worth watching.

The next test, with Boston's Beth Israel Deaconess Medical Center, will examine whether the system can triage patients waiting in emergency departments, where incomplete information and time pressure make decisions much harder.

The question is whether that impressive accuracy holds up with actual patients.

Hackers are using AI to break into companies faster

Microsoft says a China-linked hacking group used AI agents to automate cyberattacks against hundreds of organizations, sometimes gaining privileged cloud access in under 3 hours.

The speed is concerning.

The group, tracked as Storm-0940, used an open-source AI tool called ARTEX to scan systems, attempt logins, and exploit compromised application credentials. During the ConnectWise supply chain attack, it attempted to access more than 2,250 Microsoft 365 customer environments.

In one campaign, Microsoft recorded over 151,000 failed login attempts, with activity reaching 2,900 attempts per minute.

I think the most concerning detail is how quickly attackers can now move between steps that previously required significant human effort. Security teams may receive alerts on their screens while automated systems are already attempting to expand access.

Most attempts were unsuccessful, but Microsoft confirmed that attackers established privileged access in multiple organizations.

Companies may need faster detection and tighter controls over third-party applications that already have permission to access their systems.

How effectively can human security teams respond when automated attacks progress this quickly?

Prohuman team

Free Guides

Explore our free guides and products to get into AI and master it.

All of them are free to access and would stay free for you.

Feeling generous?

You know someone who loves breakthroughs as much as you do.

Share The Prohuman it’s how smart people stay one update ahead.