OpenAI is asking third-party contractors to upload real assignments and tasks from their current or previous workplaces so that it can use the data to evaluate the performance of its next-generation AI models, according to records from OpenAI and the training data company Handshake AI obtained by WIRED.

The project appears to be part of OpenAI’s efforts to establish a human baseline for different tasks that can then be compared with AI models. In September, the company launched a new evaluation process to measure the performance of its AI models against human professionals across a variety of industries. OpenAI says this is a key indicator of its progress towards achieving AGI, or an AI system that outperforms humans at most economically valuable tasks.

“We’ve hired folks across occupations to help collect real-world tasks modeled off those you’ve done in your full-time jobs, so we can measure how well AI models perform on those tasks,” reads one confidential document from OpenAI. “Take existing pieces of long-term or complex work (hours or days+) that you’ve done in your occupation and turn each into a task.”

OpenAI is asking contractors to describe tasks they’ve done in their current job or in the past and to upload real examples of work they did, according to an OpenAI presentation about the project viewed by WIRED. Each of the examples should be “a concrete output (not a summary of the file, but the actual file), e.g., Word doc, PDF, Powerpoint, Excel, image, repo,” the presentation notes. OpenAI says people can also share fabricated work examples created to demonstrate how they would realistically respond in specific scenarios.

OpenAI and Handshake AI declined to comment.

Real-world tasks have two components, according to the OpenAI presentation. There’s the task request (what a person’s manager or colleague told them to do) and the task deliverable (the actual work they produced in response to that request). The company emphasizes multiple times in instructions that the examples contractors share should reflect “real, on-the-job work” that the person has “actually done.”

One example in the OpenAI presentation outlines a task from a “Senior Lifestyle Manager at a luxury concierge company for ultra-high-net-worth individuals.” The goal is to “Prepare a short, 2-page PDF draft of a 7-day yacht trip overview to the Bahamas for a family who will be traveling there for the first time.” It includes additional details regarding the family’s interests and what the itinerary should look like. The “experienced human deliverable” then shows what the contractor in this case would upload: a real Bahamas itinerary created for a client.

OpenAI instructs the contractors to delete corporate intellectual property and personally identifiable information from the work files they upload. Under a section labeled “Important reminders,” OpenAI tells the workers to “Remove or anonymize any: personal information, proprietary or confidential data, material nonpublic information (e.g., internal strategy, unreleased product details).”

One of the files viewed by WIRED document mentions an ChatGPT tool called “Superstar Scrubbing” that provides advice on how to delete confidential information.

Evan Brown, an intellectual property lawyer with Neal & McDevitt, tells WIRED that AI labs that receive confidential information from contractors at this scale could be subject to trade secret misappropriation claims. Contractors who offer documents from their previous workplaces to an AI company, even scrubbed, could be at risk of violating their previous employers’ non-disclosure agreements, or exposing trade secrets.

“The AI lab is putting a lot of trust in its contractors to decide what is and isn’t confidential,” says Brown. “If they do let something slip through, are the AI labs really taking the time to determine what is and isn’t a trade secret? It seems to me that the AI lab is putting itself at great risk.”

While iOS users have an easy smartwatch choice in the Apple Watch, Android owners have a few more options, as well as face shapes, to choose from. The semi-squircular Samsung Galaxy Watch8 and Watch 8 Classic have both a unique look and set of health features, and are currently marked down to as low as $280 at Amazon for the 30mm Bluetooth version, or as low as $433 for the Watch8 Classic.

As the first watches to sport Google’s Wear OS 6, these made waves when they released with bigger, bolder watch faces, and an improved interface that shows more information. It has a 1.5-inch AMOLED screen that’s generously sized even on the 40mm version we tested, and has more than enough brightness to check on a sunny day.

Both watches feature the kind of physical activity and health tracking data you’d expect from a modern smartwatch, including steps, heart rate, and both sleep quality and bedtime guidance, which recommends when you should go to bed, if you couldn’t sort that out on your own. You can also use the optical sensor to measure your Antioxidant Index to help track what you’re eating without manually logging meals.

Battery life is a key factor for any smartwatch, and the smaller 40mm didn’t quite impress us, running for just right around 20 hours, about half of Samsung’s claimed runtime. The more expensive Watch8 Classic lasted closer to two full days, even some tracked physical activities tossed in. If more than full day of battery is a key selling point, I’d consider making the upgrade.

While only the graphite and silver models are in stock as I write this, there are discounts for both the 40mm and 44mm sizes across both the Bluetooth only and LTE connected options. You can also scoop a healthy markdown on the more deluxe Watch8 Classic, which I spotted for $433 in black or $450 in white. If you’re curious to learn more, make sure to check out our full review of both the Watch8 and Watch8 Classic for all the details, or peruse our other favorite smartwatches to find your new daily driver.

After creating thousands of “undressing” pictures of women and sexualized imagery of apparent minors, Elon Musk’s X has apparently limited who can generate images with Grok. However, despite the changes, the chatbot is still being used to create “undressing” sexualized images on the platform.

On Friday morning, the Grok account on X started responding to some users’ requests with a message saying that image generation and editing are “currently limited to paying subscribers.” The message also includes a link pushing people towards the social media platform’s $395 annual subscription tier. In one test of the system requesting Grok create an image of a tree, the system returned the same message.

The apparent change comes after days of growing outrage against and scrutiny of Musk’s X and xAI, the company behind the Grok chatbot. The companies face an increasing number of investigations from regulators around the world over the creation of nonconsensual explicit imagery and alleged sexual images of children. British prime minister Keir Starmer has not ruled out banning X in the country and said the actions have been “unlawful.”

Neither X nor xAI, the Musk-owned company behind Grok, has confirmed that it has made image generation and editing a paid-only feature. An X spokesperson acknowledged WIRED’s inquiry but did not provide comment ahead of publication. X has previously said it takes “action against illegal content on X,” including instances of child sexual abuse material. While Apple and Google have previously banned apps with similar “nudify” features, X and Grok remain available in their respective app stores. xAI did not immediately respond to WIRED’s request for comment.

For more than a week, users on X have been asking the chatbot to edit images of women to remove their clothes—often asking for the image to contain a “string” or “transparent” bikini. While a public feed of images created by Grok contained far fewer results of these “undressing” images on Friday, it still created sexualized images when prompted to by X users with paid for “verified” accounts.

“We observe the same kind of prompt, we observe the same kind of outcome, just fewer than before,” Paul Bouchaud, lead researcher at Paris-based nonprofit AI Forensics, tells WIRED. “The model can continue to generate bikini [images],” they say.

A WIRED review of some Grok posts on Friday morning identified Grok generating images in response to user requests for images that “put her in latex lingerie” and “put her in a plastic bikini and cover her in donut white glaze.” The images appear behind a “content warning” box saying that adult material is displayed.

On Wednesday, WIRED revealed that Grok’s standalone website and app, which is separate from the version on X, has also been used in recent months to create highly graphic and sometimes violent sexual videos, including celebrities and other real people. Bouchaud says it is still possible to use Grok to make these videos. “I was able to generate a video with sexually explicit content without any restriction from an unverified account,” they say.

While WIRED’s test of image generation using Grok on X using a free account did not allow any images to be created, using a free account on Grok’s app and website still generated images.

The change on X could immediately limit the amount of sexually explicit and harmful material the platform is creating, experts say. But it has also been criticized as a minimal step that acts as a band-aid to the real harms caused by nonconsensual intimate imagery.

“The recent decision to restrict access to paying subscribers is not only inadequate—it represents the monetization of abuse,” Emma Pickering, head of technology-facilitated abuse at UK domestic abuse charity Refuge, said in a statement. “While limiting AI image generation to paid users may marginally reduce volume and improve traceability, the abuse has not been stopped. It has simply been placed behind a paywall, allowing X to profit from harm.”