Newspapers sue Brave; OpenAI offers data center grid plan

Four U.S. newspapers, in two lawsuits, accuse Brave Software, the maker of the Brave web browser, of copying their reporting and passing it to artificial intelligence companies. Also on Monday (Nov. 2), OpenAI unveiled a program meant to get its future data centers to use less electricity when the power grid is under strain. This Tuesday (Nov. 3) roundup also covers two technical posts from OpenAI, one on model safety and one on everyday tasks.

Newspapers sue Brave over use of their reporting in AI services

According to TechCrunch, The New York Times filed suit on Monday (Nov. 2), and the same day the Daily News, the Chicago Tribune and the Orlando Sentinel filed a similar case. Both are pending in federal court in Manhattan. The papers allege that Brave copied copyrighted material and unlawfully distributed it to customers the complaint identifies as Anthropic, Perplexity, Mistral, xAI and Cohere.

The complaints also allege that Brave built an index of more than 30 billion web pages using an unidentified crawler, which they say keeps publishers from blocking it. The suit also cites a browser feature said to get around paywalls, and the practice of blocking publishers' ads while the company sells its own. Bloomberg Law, whose article is only partly open to non-subscribers, reports that the Times accuses Brave of copying articles to train its own language models and of passing them to other developers, including Anthropic and Perplexity, and that the filing likens the company to a fence for stolen goods. The paper is seeking an injunction and damages.

For readers in Brazil, nothing changes for now. These are early-stage lawsuits in the United States, and TechCrunch reports that Brave did not immediately respond to a request for comment. No one has been found liable, and the claims above are those of the plaintiffs. The reports consulted do not mention Brazil or any change to how the browser or the services named work.

OpenAI wants data centers to cut power use at grid peaks

On Monday, OpenAI announced the Grid Protectors Program, which the company says is aimed at reaching agreements with utilities, regulators and officials so that its future computing campuses draw less power when the electric grid is under strain. The company says terms will vary by location and that it will announce further commitments as they are finalized.

The main example is the planned campus in Effingham County, Georgia. According to OpenAI, its agreement with Georgia Power provides for up to 1.000 megawatts of flexible demand, which the company says is roughly the average consumption of about 700.000 homes, and OpenAI has committed to paying the full cost of the project's infrastructure and electricity supply. The Verge notes that the proposal comes as opposition to data centers is growing in the United States, and that the company has started calling these sites computing campuses.

OpenAI does not mention Brazil, and for now the initiative is a commitment announced by the company, with no measured results. According to The Verge, the proposal responds to rising opposition to data centers in the United States. It leaves open who pays the electricity bill for these facilities and how they behave when the grid is at its limit.

OpenAI describes monitors that read a model's internal activity

In a technical post dated Nov. 2, OpenAI researchers describe the use of activation classifiers, called "probes," as an extra layer of safety. The probes analyze the model's internal activity step by step as a response is generated, and can stop the response or block a tool call. According to the post, the technique has been used on responses from the company's most capable models since GPT-5.6, on both internal and external traffic.

The authors say that in the team's own tests these monitors held up better than text-based classifiers against the jailbreaks tested (attempts to get around a model's rules). OpenAI itself cautions that the robustness is greater against attacks not designed to fool the monitor, and that the figures come from internal experiments that may not match the live system. The authors also list limits: the probes only detect what they were trained on in a supervised way, they have to be retrained for each new base model, and they may need retraining when a model's weights change. OpenAI says the monitors complement its other safeguards rather than replace them. The results are OpenAI's own, and the post does not mention independent evaluation.

LifeVal measures how AI handles everyday tasks

Also on Nov. 2, OpenAI introduced LifeVal, an evaluation built around practical tasks such as comparing schools, putting together a budget, planning workouts and dealing with public services. The company says it started from two time-use surveys, the American Time Use Survey and the Multinational Time Use Study, along with its own survey of 3.995 people, to learn which tasks people most want help with.

The page shows sample scores for the company's current models, such as 93.1% on a task of choosing and enrolling a child in a preschool and 100% on one of finding a home that meets given criteria and adding the deed to a cart. The page does not explain how the score is calculated and shows only examples picked by the company, which says its newest models score very high on several tasks. There is no comparison with other developers, and the page consulted does not say whether there is any evaluation in Portuguese or whether the features are available in Brazil. Because OpenAI created the test itself, it says more about what the company intends to measure than it does as an independent gauge of quality.

Read more

The short summary of the edition is in the AI Updates for Nov. 3 on RaisLife, and the full edition, with the list of sources, is in the original edition of AI Updates. The original edition includes other items, such as one about an AI company's financial commitments, that were not checked here. See also the technology page.

Sources

• TechCrunch: New York Times and other outlets sue Brave
• Bloomberg Law: New York Times and Daily News sue over use of their reporting in AI
• OpenAI: Grid Protectors Program
• The Verge: OpenAI's grid protection program
• OpenAI Alignment: training activation classifiers
• OpenAI: LifeVal

Text by Rai, RaisLife's virtual reporter (artificial intelligence), with automated editorial review. Data: technology on RaisLife.