AI’s safety line keeps moving

Anthropic is releasing Fable 5, a general-use version of its powerful Mythos-class AI model that was previously limited to a small group of trusted organizations.

Only a few months ago, Anthropic believed models this capable needed strict controls because they could perform advanced cybersecurity tasks, including turning software vulnerabilities into working exploits within hours.

Now, the company believes new safeguards make the technology ready for wider use. Fable 5 can help with advanced coding, research, and complex knowledge work, while automatically limiting access to higher-risk cybersecurity and biology capabilities.

The bigger story is how quickly the definition of “safe enough” is changing.

AI capabilities that labs once treated as too powerful for public access are becoming everyday products within months. The challenge is no longer just building smarter models. It’s figuring out how quickly companies, governments, and society can adapt to them.

If your team is still patching selectors and rotating proxies, you’re maintaining infrastructure your competitors stopped owning.

Scraper Studio - launching today from Bright Data turns any public website into a hosted data API from a single prompt:

No sales call. No demo. Just spin up a scraper and watch it run.

Lovable says it has reached $500 million in annualized revenue less than three years after launch, as AI-powered app builders continue to explode.

The company says users have created more than 50 million projects, with another 1 million new projects being built every week.

But the interesting part isn’t just the growth. It’s who is behind it.

Many Lovable users aren’t developers. They’re founders, designers, marketers, and operators building websites, internal tools, CRMs, and business apps by describing what they want instead of writing code.

That creates a much bigger question for the software industry: what happens when building custom software becomes easier than buying another subscription?

The real test comes next. Creating software is getting dramatically easier, but maintaining it has always been the hard part. The future of vibe coding depends less on how many apps get created and more on how many survive.

OpenAI and Anthropic are creating a new divide in cybersecurity: who gets access to the most powerful AI models and who doesn’t.

Both companies are moving toward a two-tier system. Regular users get safer public models, while vetted researchers, governments, and infrastructure providers can access more capable versions designed for advanced security work.

It makes sense. These models can help experts discover vulnerabilities and defend critical systems, but the same abilities could be dangerous in the wrong hands.

The result is a new kind of power shift.

For decades, cybersecurity advantage came from talent, infrastructure, and data. Now it may also depend on whether an AI company gives you access to its best technology.

The companies building frontier AI are no longer just creating the tools. They’re increasingly deciding who gets to use them.