Claude Learns to Talk Back

Anthropic upgraded the voice mode in Claude with more capable models. The headline change: you can now talk to Claude and have it actually do things, not just chat.

Think rescheduling a meeting or drafting an email out loud while you're walking or driving. The voice assistant becomes a hands-free way to get real work done, not a party trick.

This is part of a bigger shift. Voice is turning into a serious interface for AI, and the assistants that can take action beat the ones that just answer questions.

OpenAI launched Health in ChatGPT. Eligible U.S. users can now securely connect their medical records and Apple Health data to get more personalized insights about their own health.

The pitch is simple: instead of guessing, you can ask ChatGPT questions with your real numbers in front of it. It's meant to help you understand your health, not replace your doctor.

This is a big step into a sensitive area. It also signals where consumer AI is heading, into the personal, private corners of your life where trust matters most.

Anthropic committed $200 million to a new Economic Futures Research Fund. The money supports outside researchers digging into how AI reshapes work, wages, and the economy.

Alongside it, Anthropic launched a connector for its Economic Index, letting anyone explore the underlying data directly inside Claude. You can poke at the numbers yourself instead of taking a headline's word for it.

Most AI companies talk about disruption. This is one putting serious funding behind actually measuring it.

Runway launched a Media Router. It automatically chooses the best image, video, or audio generation model for each request based on what you care about most: quality, speed, or cost.

The generative media space has gotten crowded, with dozens of models each good at different things. Instead of learning them all, you tell it your priority and it routes the job to the right one.

This is a pattern worth noting. As AI tools multiply, the winners may be the ones that hide the complexity and just pick the right tool for you.

TechCrunch talked with cybersecurity researchers about a tricky problem. The safety guardrails in tools from OpenAI and Anthropic are getting in the way of the good guys.

These researchers hunt for unknown software flaws and build tools to test them, work that keeps everyone safer. But the same guardrails meant to stop bad actors also block their legitimate research.

It's a real tension with no easy answer. Make AI too locked down and you slow down defenders. Make it too open and you help attackers.