Bonta says CA will investigate Grok’s sexually explicit deepfakes

“@grok put her in a bikini”

Writing nothing but a short prompt, users on the social media platform X can quickly generate nonconsensual and sexually explicit images of women and children — a disturbing trend that has prompted California to launch a probe into X’s artificial intelligence tool.

As CalMatters’ Khari Johnson explains, California Attorney General Rob Bonta on Wednesday said the state plans to investigate the AI chatbot Grok after reports that a recent update enables users to create and spread sexualized photos of women and children. Tech billionaire Elon Musk’s company xAI developed Grok. 

• Bonta, in a statement: “The avalanche of reports … in recent weeks is shocking. This material, which depicts women and children in nude and sexually explicit situations, has been used to harass people across the internet.”

In a Jan. 3 post on X, Musk said that anyone using the AI tool to make illegal content “will suffer the same consequences as if they upload illegal content.” Though, in an earlier response to Reuters addressing reports of sexualized images of kids spreading on X, he said that “Legacy Media Lies.”

Since 2019, California has passed about a half a dozen laws protecting people from deepfake pornography. Most recently, it enacted a law that punishes people providing “a service that enables the ongoing operation of deepfake pornography” with a fine of up to $250,000.

• Assemblymember Rebecca Bauer-Kahan, a San Ramon Democrat and bill author, in a statement to CalMatters: “Real women are having their images manipulated without consent, and the psychological and reputational harm is devastating. Underage children are having their images used to create child sexual abuse material, and these websites are knowingly facilitating it.”

With the help of Grok — which will also be soon integrated in the Pentagon’s systems — X now produces more nonconsensual naked or sexual imagery than any other website online, according to Bloomberg. And while the images may be fake, the damage they inflict is real: In 2024, the FBI warned that the use of deepfake tools to extort young people has led to self harm and suicide. 

Read more here.

From CalMatters politics reporter Maya C. Miller:

A panel of three federal court judges shot down the latest “Hail Mary” attempt by California Republicans and President Donald Trump’s administration to halt the use of the new congressional maps voters approved under Proposition 50, saying their “molehill of statements” claiming the maps were racially biased did not make that claim true.

The 2-1 ruling, with one Trump-appointed judge dissenting, comes after the judges heard three days of testimony and reviewed more than 500 exhibits totaling thousands of pages along with audio and video recordings.

“Our conclusion probably seems obvious to anyone who followed the news in the summer and fall of 2025,” wrote Judge Josephine Staton, an Obama appointee, in her scathing rebuke of the GOP’s challenge. 

She noted that opponents of Prop. 50 relied on a completely different argument against the measure in court than they used on the campaign trail.

• Staton, in the ruling: “So, Challengers have abandoned the argument they made to the voters. Proposition 50, apparently, is no longer a partisan power grab. Now, it is a ‘racial gerrymander.’”

The GOP said they will seek an injunction from the U.S. Supreme Court. 

Fewer California prisoners are being found suitable for parole, despite the state expanding opportunities and incentives for incarcerated people to seek parole over the last 10 years, reports CalMatters’ Joe Garcia.

Before plateauing in 2023, the annual number of parole hearings increased over the years, from 5,226 in 2018 to 9,017 in 2022. But the success rate for prisoners to be found suitable for parole has declined, from 39% in 2018 to below 25% in 2025.

Parole experts said various factors could be at play. As California’s prison population ages, the mental and physical limitations of these older incarcerated people could negatively impact their ability to present themselves before a parole board.

More than ever before, parole commissioners also have an abundant amount of personal information on individual prisoners: Visitor logs, personal expenses, medical records and monitored phone calls and text messages can all be deemed relevant when considering parole. The extra scrutiny could reveal potential misconduct, and result in commissioners granting fewer paroles.

Read more here.

Bonta, California’s chief prosecutor, spoke with CalMatters to discuss the state’s recent legal clashes with the Trump administration, the federalization of National Guard troops and preparing for the next potential large-scale federal immigration raids. Read more from CalMatters’ Nigel Duara.