Technologies

OpenAI and Google Take Steps to Avoid Abusive AI Imagery After Grok Scandal

AI safety, especially around images and videos, continues to be an evolving challenge.

2026 started with a horrifying example of generative AI’s potential for abuse. Grok, the AI tool from Elon Musk’s xAI, was used to undress or nudify pictures of people shared on X (formerly Twitter) at an alarming rate. Grok made 3 million sexualized images over a span of 11 days in January, with approximately 23,000 of those containing images of children, according to a study from the Center for Countering Digital Hate.

Now, competitors like OpenAI and Google are stepping up their security to avoid being the next Grok.

Advocates and safety researchers have long been concerned about AI’s ability to create abusive and illegal content. The creation and sharing of nonconsensual intimate imagery, sometimes referred to as revenge porn, was a big problem before AI. Generative AI only makes it quicker, easier and cheaper for anyone to target and victimize people.

On Jan. 14, two weeks into the scandal, X’s Safety account confirmed in a post that it would pause Grok’s ability to edit images on the social media app. Grok’s image-generation abilities are still available to paying subscribers in its standalone app and website. X did not respond to multiple requests for comment.

Most major companies have safeguards in place to prevent the kind of wide-scale abuse that we saw was possible with Grok. But cybersecurity is never a solid metal wall of protection; it’s a brick wall that’s constantly undergoing repairs. Here’s how OpenAI and Google have tried to beef up their safety protections to circumvent Grok-like failures.

OpenAI fixes image generation vulnerabilities

At a base level, all AI companies have policies prohibiting the creation of illegal imagery, like child sexual abuse material, also known as CSAM. Many tech companies have guardrails to prevent the creation of intimate imagery altogether. Grok is the exception, with «spicy» modes for image and video.

Still, anyone intent on creating nonconsensual intimate imagery can try to trick AI models into doing so.

Researchers from Mindgard, a cybersecurity company focused on AI, found a vulnerability in ChatGPT that allowed people to circumvent its guardrails and make intimate images. They used a tactic called «adversarial prompting,» where testers try to poke holes in an AI with specifically crafted instructions. In this case, it was tricking the chatbot’s memory with custom prompts, then copying the nudified style onto images of well-known people.

Mindgard alerted OpenAI of its findings in early February, and the ChatGPT developer confirmed on Feb. 10 — before Mindgard went public with its report — that it had fixed the problem.

«We’re grateful to the researchers who shared their findings,» an OpenAI spokesperson said to CNET and Mindgard. «We moved quickly to fix a bug that allowed the model to generate these images. We value this kind of collaboration and remain focused on strengthening safeguards to keep users safe.»

This process is how cybersecurity often works. Outside red-team researchers like Mindgard test software for weaknesses or workarounds, mimicking strategies that bad actors might use. When they identify security gaps, they alert the software provider so fixes can be deployed.

«Assuming motivated users will not attempt to bypass safeguards is a strategic miscalculation. Attackers iterate. Guardrails must assume persistence,» Mindgard wrote in a blog post.

While tech companies boast about how you can use their AI for any purpose, they also need to make a strong promise that they can prevent AI from being used to enact abuse. For AI image generation, that means having a strong repertoire of prompts that will be refused and kicked back to users.

When OpenAI launched its Sora 2 video model, it promised to be more conservative with its content moderation for this very reason. But it’s important to ensure its moderation practices are consistently effective, not just at a product’s launch. It makes AI safety testing an ongoing process for cybersecurity researchers and AI developers alike.

Google upgrades Search reporting

For its part, Google is taking steps to ensure abusive images aren’t spread as easily. The tech giant simplified its process for requesting the removal of explicit images from Google Search. You can click the three dots in the upper right corner of an image, click report and then tell Google you want the photo removed because it «shows a sexual image of me.» The new changes also let you select multiple images at once and track your reports more easily.

«We hope that this new removal process reduces the burden that victims of nonconsensual explicit imagery face,» the company said in a blog post.

When asked about any further steps the company is taking to prevent AI-enabled abuse, Google pointed CNET to its generative AI prohibited use policy. Google’s policy, like many other tech companies’ fine print, outlaws using AI for illegal or potentially abusive activities, such as creating intimate imagery.

There are laws that aim to help victims when these images are shared online, such as the 2025 Take It Down Act. But that law’s scope is limited, which is why many advocacy groups, like the National Center on Sexual Exploitation, are pushing for better rules.

There’s no guarantee that these changes will prevent anyone from ever using AI for harassment and abuse. That’s why it’s so important that developers stay vigilant to ensure we are all protected — and act quickly when reports and problems pop up.

(Disclosure: Ziff Davis, CNET’s parent company, in 2025 filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)

Technologies

Google races to put Gemini at the center of Android before Apple’s AI reboot

Google is using its latest Android rollout to position Gemini as the AI layer across phones, Chrome, laptops and cars.

Google is using its latest Android rollout to make Gemini less of a chatbot and more of an operating layer across the phone, browser, car and laptop, just weeks before Apple is expected to show its own Gemini-powered Apple Intelligence reboot at WWDC.
Ahead of its Google I/O developer conference next week, the company previewed a number of Android updates, including AI-powered app automation, a smarter version of Chrome on Android, new tools for creators, a redesigned Android Auto experience, and a sweeping set of new security features.
Alphabet is counting on Gemini to help Google compete directly with OpenAI and Anthropic in the market for artificial intelligence models and services, while also serving as the AI backbone across its expansive portfolio of products, including Android. Meanwhile, Gemini is powering part of Apple’s new AI strategy, giving Google a role in the iPhone maker’s reset even as it races to prove its own version of personal AI on the phone is further along.
Sameer Samat, who oversees Google’s Android ecosystem, told CNBC that Google is rebuilding parts of Android around Gemini Intelligence to help users complete everyday tasks more easily.
“We’re transitioning from an operating system to an intelligence system,” he said.
As part of Tuesday’s announcements. Google said Gemini Intelligence will be able to move across apps, understand what’s on the screen and complete tasks that would normally require a user to jump between multiple services. That means Android is moving beyond the traditional assistant model, where users ask a question and get an answer, and acting more like an agent.
For instance, Google says Gemini can pull relevant information from Gmail, build shopping carts and book reservations. Samat gave the example of asking Gemini to look at the guest list for a barbecue, build a menu, add ingredients to an Instacart list and return for approval before checkout.
A big concern surrounding agentic AI involves software taking action on a user’s behalf without permissions. Samat said Gemini will come back to the user before completing a transaction, adding, “the human is always in the loop.”
Four months after announcing its Gemini deal with Google, Apple is under pressure to show a more capable version of Apple Intelligence, which has been a relative laggard on the market. Apple has long framed privacy, hardware integration and control of the user experience as its advantages.
Google’s Android push is designed to show it can bring AI deeper into the device experience while still giving users control over what Gemini can see, where it can act and when it needs confirmation.
The app automation features will roll out in waves, starting with the latest Samsung Galaxy and Google Pixel phones this summer, before expanding across more Android devices, including watches, cars, glasses and laptops later this year.
The company is also redesigning Android Auto around Gemini, turning the car into another major surface for its assistant. Android Auto is in more than 250 million cars, and Google says the new release includes its biggest maps update in a decade and Gemini-powered help with tasks like ordering dinner while driving.
Alphabet’s AI strategy has been embraced by Wall Street, which has pushed the company’s stock price up more than 140% in the past year, compared to Apple’s roughly 40% gain. Investors now want to see how Gemini can become more central to the products people use every day.
WATCH: Alphabet briefly tops Nvidia after report of $200 billion Anthropic cloud deal

Technologies

Waymo recalls 3,800 robotaxis after glitch allowed some vehicles to ‘drive into standing water’

Waymo issued a voluntary recall of about 3,800 of its robotaxis to fix software issues that could allow them to drive into flooded roadways.

Waymo is recalling about 3,800 robotaxis in the U.S. to fix software issues that could allow them to “drive onto a flooded roadway,” according to a letter on the National Highway Traffic Safety Administration’s website.
The voluntary recall is for Waymo vehicles that use the company’s fifth and sixth generation automated driving systems (or ADS), the U.S. auto safety regulator said in the letter posted Tuesday.
Waymo autonomous vehicles in Austin, Texas, were seen on camera driving onto a flooded street and stalling, requiring other drivers to navigate around them. It’s the latest example of a safety-related issue for the Alphabet-owned AV unit that’s rapidly bolstering its fleet of vehicles and entering new U.S. markets.
Waymo has drawn criticism for its vehicles failing to yield to school buses in Austin, and for the performance of its vehicles during widespread power outages in San Francisco in December, when robotaxis halted in traffic, causing gridlock.
The company said in a statement on Tuesday that it’s “identified an area of improvement regarding untraversable flooded lanes specific to higher-speed roadways,” and opted to file a “voluntary software recall” with the NHTSA.
“Waymo provides over half a million trips every week in some of the most challenging driving environments across the U.S., and safety is our primary priority,” the company said.
Waymo added that it’s working on “additional software safeguards” and has put “mitigations” in place, limiting where its robotaxis operate during extreme weather, so that they avoid “areas where flash flooding might occur” in periods of intense rain.
WATCH: Waymo launches new autonomous system in Chinese-made vehicle

Technologies

Qualcomm tumbles 13% as semiconductor stocks retreat from historic AI-fueled surge

Semiconductor equities reversed sharply after a broad AI-driven advance, with Qualcomm suffering its worst day since 2020 amid inflation concerns and rising oil prices.

Semiconductor stocks fell sharply on Tuesday, reversing course after an extensive rally that had expanded the artificial intelligence investment theme well past Nvidia and driven the industry to unprecedented levels.

Qualcomm plunged 13% and was on track for its steepest single-day decline since 2020. Intel shed 8%, while On Semiconductor and Skyworks Solutions each lost more than 6%. The iShares Semiconductor ETF, which benchmarks the overall sector, fell 5%.

The sell-off came after a key gauge of consumer prices came in above forecasts, and as conflict in Iran pushed crude oil higher—prompting investors to shift away from riskier assets.

The preceding advance had widened the AI opportunity set beyond longtime industry leader Nvidia, which for much of the past several years had largely carried the market to new peaks on its own.

Explosive appetite for central processing units, along with the graphics processing units that power large language models, has sent chipmakers to all-time highs.

Market participants are wagering that the shift from AI model training to autonomous agents will lift demand for additional AI hardware. Among the beneficiaries are memory chip producers, which are raising prices as supply remains tight.

Micron Technology slid 6%, and Sandisk cratered 8%. Sandisk’s stock has surged more than six times over since January.