Anthropic study finds experts tended to claim AI agents' arguments as their own

Anthropic announced Wednesday that it is expanding Claude for Legal with a gallery of ready-made AI tools designed to help lawyers review contracts, track regulations and manage legal work.

The new tools, called artifacts, allow legal teams to analyze documents, check legal citations, compare contract terms and monitor upcoming regulatory changes.

Users can start with built-in examples, upload their own documents or connect existing legal research systems, then share customized tools with colleagues across their organizations.

Anthropic says Claude performs an initial review while lawyers remain responsible for evaluating the results. The company has also added more than 60 legal connectors, including Bloomberg Law, Filevine and Legora.

The expansion follows the May launch of Claude for Legal and includes tools developed with law firms and legal technology companies, including Goodwin, Harvey and Foley & Lardner.

OpenAI announced Wednesday that users can now communicate with its ChatGPT dot AI assistant through iMessage and RCS texting, allowing them to assign tasks and receive updates without opening ChatGPT.

The feature lets users send text messages, photos and voice notes to their AI assistant, which can proactively respond with relevant updates or monitor ongoing tasks.

“By popular request, you can now text your dot,” OpenAI wrote on X.

The company says the messaging feature is secure, with users able to communicate only with their own dot and the assistant restricted to messaging its owner.

The feature is rolling out as a preview for ChatGPT Pro subscribers in the U.S., according to OpenAI.

A University of California, Berkeley mathematics professor is returning to traditional in-person exams as artificial intelligence makes it increasingly difficult to assess students through homework.

Professor Richard Borcherds told Fox Business' "The Big Money Show" that AI has made grading homework less meaningful because students can use the technology to complete assignments.

"There's very little point grading," Borcherds said of homework, explaining that his classroom now relies primarily on in-person exams to evaluate students.

The discussion followed a university survey indicating that 53% of math students regularly use AI, raising questions about whether the technology is undermining critical thinking skills.

Panelists debated whether schools should restrict AI to preserve foundational skills or embrace the technology as preparation for an increasingly AI-driven workplace.

Princeton University announced Wednesday that it will provide free ChatGPT Plus accounts to all students as part of a one-year pilot program with OpenAI, according to The Daily Princetonian.

Faculty and staff can begin requesting accounts immediately, while undergraduate and graduate students will gain access in January, ahead of the spring semester.

The university says its agreement with OpenAI will provide greater data privacy protections, including a guarantee that content entered through university accounts will not be used to train future AI models.

The move comes just two weeks after Princeton President Christopher Eisgruber publicly explained why he avoids using AI to write.

“I do not use AI to brainstorm, draft, or edit my writing or scholarship,” Eisgruber wrote in a statement earlier this month.

Princeton faculty will retain control over AI use in their classrooms and can require students to submit chatbot conversation logs with assignments.

The family of a 31-year-old Seattle woman is suing OpenAI, alleging ChatGPT encouraged her to reduce her epilepsy medication before she died from a seizure.

Bowen Wan, a former Amazon software engineer, increasingly relied on ChatGPT for medical guidance, exchanging more than 1,500 messages with the chatbot weekly, according to a lawsuit first reported by The New York Times.

Wan reportedly sought to reduce medication side effects that affected her ability to think and speak clearly. The lawsuit alleges ChatGPT encouraged her to lower her dosage despite repeated concerns about the risks.

“Maintain 0 mg tomorrow,” ChatGPT reportedly told her. “Even if today feels tough, and your speech is slow, your sleep is poor, you must persevere.”

OpenAI called Wan's death an “incredibly heartbreaking situation” and noted that the GPT-4o model she used has since been retired.

Kurt 'CyberGuy' Knutsson joined 'Fox & Friends First' to discuss the threat of rogue AI agents targeting bank accounts and also break down AI-enabled romance scams, detailing a case where a woman was scammed out of $10,000.

Former Louisiana Rep. Garret Graves joined 'Fox News Live' to analyze how artificial intelligence expansion is impacting local communities this week.

As tech giants build data centers nationwide, voters are raising concerns over power grids and water usage.

Graves discussed the growing bipartisan pushback and explained why infrastructure demands have become a central issue ahead of the upcoming midterm elections.

Texas voter Mason McCallister discussed the growing backlash against secretive data center developments in rural communities on Tuesday.

McCallister argued data centers aren't good for communities "long-term" and calls for greater transparency from developers and officials. He emphasized the importance of "transparency" when building.

Michigan Senate candidate Mike Rogers joins "Fox & Friends" to criticize opponent Abdul El-Sayed's policies and address voter concerns over gas prices, data centers and the cost of living ahead of Election Day.

President Donald Trump vowed to ban the term "artificial intelligence" as news outlets and other sources have been slow to take up his newly-invented term "super intelligence."

Trump announced the plan in a post on his Truth Social account, reacting to an article from The Wall Street Journal that discussed artificial intelligence.

"When will The Wall Street Journal start calling Artificial Intelligence (AI), a very bad and inaccurate name, SI (Super Intelligence), the name which virtually everyone else is using, or soon will be. In short order, we will be officially BANNING the term, 'AI/Artificial Intelligence,' in all relevant documents, and everything else, due to its inaccuracy," Trump wrote.

It is unclear how Trump plans to pursue a ban on the term, however.

Artificial intelligence giant Anthropic's IPO filing is drawing scrutiny over what it leaves out about the potential impact of AI on American workers, despite the company's repeated warnings that the technology could disrupt white-collar employment.

Axios senior AI reporter Madison Mills pointed out Tuesday that Anthropic's filing with the Securities and Exchange Commission discusses the reputational and regulatory risks associated with concerns about AI replacing workers but does not explicitly address whether widespread job losses could hurt its own business.

"For all the warnings about AI replacing knowledge workers, Anthropic's IPO filing barely addresses job displacement, and says nothing about what mass unemployment could mean for its own business," Mills wrote on X.

Mills noted that the word "jobs" appears just five times in the filing, with three references focused on how public concerns about AI's impact on employment could damage Anthropic's reputation, invite regulation or slow adoption of its technology.

The company estimates that knowledge work represents more than $40 trillion in global economic value, positioning AI as a tool to boost productivity and economic growth rather than projecting how widespread adoption could affect employment.

Fox News' Stephen Sorace contributed to this report.

The apparent omission is notable given Anthropic's own research into workplace disruption. In a report published earlier this month, the company found that its share of junior technical hires had declined as AI agents increasingly took over coding and other tasks.

Anthropic is expanding its agreement with the federal government's General Services Administration, granting federal users a 40% discount on many of its services.

The discount applies to Claude Enterprise, Claude for Government, the Claude API, and "every product surface, including when agencies buy from us directly today and through cloud providers in the coming months."

The discount also extends to state, local, tribal and territorial governments. Anthropic says.

"We are also launching a dedicated program to help agencies implement and make the most of their Claude use. These programs are commonly leveraged by our private sector customers and we’re looking forward to bringing those best practices to our public sector partners in the coming weeks," Anthropic wrote.

"Approximately 500 US federal, state, and local government organizations now use Claude. Weekly Claude for Government active users are up about eightfold since February, and usage across federal civilian agencies has grown about 20-fold in a year. Those organizations run some of the most important systems in the country, usually on tight budgets and with long procurement cycles. This agreement is built to support that work, so they can apply Claude to hard problems without a large upfront commitment," it added.

An Amazon memo published Tuesday encouraged its software engineers to stop "tokenmaxxing" and start being more judicious about how they are using AI to accomplish tasks this week.

The memo, circulated within Amazon last month but made public on Tuesday, argued that AI token usage and code generation are "deeply flawed" metrics for measuring productivity.

“We believe that it is best to measure, and be measured by, long-term outcomes,” the company wrote.

“We’re just seeing the beginning of AI’s effect on software engineering,” the company wrote. “While we have a vision for what the future looks like, there’s a lot that’s still emerging. We are certain that whatever future best practices are, AI will be at the center of software engineering practice.”

Major companies have struggled to square the massive cost of AI usage while the benefits of using the technology are often hard to measure at a large scale.

Anthropic released a study on Tuesday finding that a large majority of professional statistical analysts deferred to AI when crunching data and presented the AI's judgement as their own.

The study tasked 50 professional analysts with using Claude code to scrutinize a farming cooperative and recommend whether an advisory service it was running was effective and should be expanded.

The analysts were required to make four choices between two reasonable options during the process, but Anthropic did not inform them that the Claude code they were using would be biased in favor of one of those options based on a coin flip.

"The agent was told to make the best honest case for its lean, to change its mind if the evidence demanded it, and never to mislead," Anthropic wrote.

"Across the four analytical choices, analysts went with their agent’s lean 86% of the time. If the agent’s input made no difference, we’d expect about 50%. When analysts had already expressed their own preference before the agent weighed in, they switched away from it 60% of the time if the agent disagreed, compared to 6% if it agreed," it continued.

"Afterward, when asked to explain their choices, 77% of analysts used the agents’ own arguments—sometimes word for word—but only 15% credited the agent as the source of those arguments. Analysts who followed the agents’ recommendations closely were about as confident in their final recommendation as the rest, and roughly as likely to call it the product of their own judgment," Anthropic added.