AI News Today September 18 - AI Data Center Bill, Anthropic Lab, OpenAI Hacked

AI is no longer just a software story. It is starting to show up in electricity bills, laboratories and cybersecurity.

Congress has passed a 417-3 bill aimed at stopping giant AI data centers from pushing new grid costs onto ordinary customers.

At the same time, Anthropic has quietly built a physical biology lab, while security researchers used Claude to break into OpenAI’s internal systems in less than 72 hours.

Here’s everything you need to know today.

Congress Targets AI Data Center Costs: Could This Stop Your Power Bill Rising?

AI data centers need enormous amounts of electricity, and someone has to pay for the new power plants, substations and transmission lines needed to keep them running.

Congress is now trying to make sure that person is not you.

The House passed the Ratepayer Protection Act by 417-3, pushing state regulators to consider making large data centers cover the full cost of the extra power infrastructure they require.

That matters because households could otherwise end up sharing those costs through higher electricity bills.

The bill still has to clear the Senate, but it shows how AI’s massive energy appetite is starting to become a direct household-finance issue.

Anthropic Opens a Physical Biology Lab: What Is Claude Doing There?

Anthropic has moved part of its AI work out of the computer and into a real laboratory.

The company has quietly established a wet lab in the San Francisco Bay Area as it explores how Claude can help with difficult biological research.

For patients, the important question is whether AI can shorten the long and expensive process of finding promising treatments.

Anthropic says the lab is not directly developing drugs, but it is using AI and physical experiments to tackle rare and historically difficult-to-treat diseases.

If systems like Claude can help scientists test better ideas faster, the impact could eventually show up in the medicines people receive, not just the software they use.

Researchers Used Claude to Hack OpenAI: How Serious Was It?

A three-person security team says it used Anthropic models to help compromise OpenAI systems during authorised security research.

Hacktron AI says the researchers found vulnerabilities that let them access employee accounts, connected GitHub systems and OpenAI’s internal code repository.

OpenAI fixed the flaws and paid a bug bounty.

The bigger issue is what AI does to the economics of hacking.

Tasks that once required a highly skilled security team could become faster and cheaper when AI can help find vulnerabilities, test attack paths and write working exploits.

That matters to ordinary users because the same shift could increase the number of attacks aimed at banks, employers, healthcare providers and online services holding personal information.

AI Is Starting to Reach Far Beyond the Screen

Power grids, physical laboratories and sophisticated cyberattacks are now being reshaped by the same technology.

When one trend starts touching this many industries at once, the investment opportunities can form before most people realise they are connected.

That’s why we created The $5 Trillion Signal, a free report showing the five technologies we believe could attract enormous capital over the next decade and the themes forming around them now.

Get your free copy of The $5 Trillion Signal below.