OctoML says it has ended its business relationship with Civitai days after an investigation by 404 Media revealed the text-to-image platform was being used to generate images that “could be categorized as child pornography.” Civitai previously relied on OctoML’s OctoAI for image generation. While OctoML initially indicated it would continue working with Civitai and introduced new measures to curb the creation of harmful images, 404 Media reported on Saturday that it has now decided to cut ties with the platform altogether.

According to 404 Media’s December 5 report, internal communications showed that OctoML was aware some Civitai users were creating sexually explicit material that included nonconsensual images of real people and pornographic depictions of children. In a followup report this weekend, the publication noted that OctoML rolled out a filter to block the generation of all NSFW content on Civitai before announcing its decision to pull out. Civitai also added new moderation methods in response to the investigation earlier this week, including a mandatory embedding called Civitai Safe Helper (Minor) that bars the model from generating images of children if “a mature theme or keyword is detected,” according to 404.

Andreessen Horowitz-backed Civitai has previously come under scrutiny for its “bounties” feature that challenges users to generate realistic images of real people for rewards. In November, 404 Media found it was being used to make nonconsensual deepfakes of celebrities, influencers and even a private citizen that were, in many cases, sexual in nature. The subjects were primarily women. According to 404 Media, Civitai has since added a filter to prevent the creation of NSFW content featuring “certain” celebrities.

Nevertheless, OctoML — which uses Amazon Web Services’ servers — no longer wants to be involved. In a statement to 404 Media on Saturday, OctoML said, “We have decided to terminate our business relationship with Civitai. This decision aligns with our commitment to ensuring the safe and responsible use of AI.”

This article originally appeared on Engadget at https://www.engadget.com/controversial-ai-image-platform-civitai-has-been-dropped-by-its-cloud-computing-provider-195530538.html?src=rss

OctoML says it has ended its business relationship with Civitai days after an investigation by 404 Media revealed the text-to-image platform was being used to generate images that “could be categorized as child pornography.” Civitai previously relied on OctoML’s OctoAI for image generation. While OctoML initially indicated it would continue working with Civitai and introduced new measures to curb the creation of harmful images, 404 Media reported on Saturday that it has now decided to cut ties with the platform altogether.

According to 404 Media’s December 5 report, internal communications showed that OctoML was aware some Civitai users were creating sexually explicit material that included nonconsensual images of real people and pornographic depictions of children. In a followup report this weekend, the publication noted that OctoML rolled out a filter to block the generation of all NSFW content on Civitai before announcing its decision to pull out. Civitai also added new moderation methods in response to the investigation earlier this week, including a mandatory embedding called Civitai Safe Helper (Minor) that bars the model from generating images of children if “a mature theme or keyword is detected,” according to 404.

Andreessen Horowitz-backed Civitai has previously come under scrutiny for its “bounties” feature that challenges users to generate realistic images of real people for rewards. In November, 404 Media found it was being used to make nonconsensual deepfakes of celebrities, influencers and even a private citizen that were, in many cases, sexual in nature. The subjects were primarily women. According to 404 Media, Civitai has since added a filter to prevent the creation of NSFW content featuring “certain” celebrities.

Nevertheless, OctoML — which uses Amazon Web Services’ servers — no longer wants to be involved. In a statement to 404 Media on Saturday, OctoML said, “We have decided to terminate our business relationship with Civitai. This decision aligns with our commitment to ensuring the safe and responsible use of AI.”

This article originally appeared on Engadget at https://www.engadget.com/controversial-ai-image-platform-civitai-has-been-dropped-by-its-cloud-computing-provider-195530538.html?src=rss

Google has canceled its Gemini launch events scheduled for next week and now plans to launch its GPT-4 competitor in January, according to The Information. The company teased Gemini at I/O 2023, touting it then as a foundational model with “impressive multimodal capabilities” early in its training.

Google initially planned to debut it before the end of the year in an unusually timed launch between the holidays — which it didn’t publicly announce — but sources told The Information that Gemini was struggling with non-English queries, prompting CEO Sundar Pichai to delay its release. Gemini is intended to be able to handle a broad range of applications, combining different types of data like images and text for more advanced tasks. In May, Google said, “Once fine-tuned and rigorously tested for safety, Gemini will be available at various sizes and capabilities.”

Gemini is expected to bring improvements to Google’s existing AI and AI-enhanced products too, like Bard, Google Assistant and Search. But, given the significance of this release and the dominance rival OpenAI already has in the space, it seems Google isn’t going to risk rushing Gemini out the door.

This article originally appeared on Engadget at https://www.engadget.com/google-is-reportedly-pushing-the-launch-of-its-gemini-ai-to-2024-173444507.html?src=rss

Amazon announced on Friday that it’s signed a contract with SpaceX to deliver batches of its Project Kuiper satellites to low Earth orbit in 2025. SpaceX is undoubtedly Amazon’s biggest competitor as it breaks into the satellite internet space, and already has a constellation of over 4,000 Starlink satellites in operation. It’s also a rival of Blue Origin, the aerospace company founded by Jeff Bezos that has its own rockets in development. But when it comes to launches, SpaceX’s pace and the reliability of its Falcon 9 rocket is unmatched.

The contract with SpaceX is for three Falcon 9 launches, Amazon said in a blog post. They’re expected to lift off in mid-2025. Amazon is planning to start customer pilots of its Project Kuiper satellite internet service by the end of next year, and will soon start deploying the fleet that will support it. It launched its first two prototype satellites in October. The company already has deals for upcoming launches on United Launch Alliance’s (ULA) Vulcan rocket, Arianespace’s Ariane 6 and Blue Origin’s New Glenn — all of which have been hit by development delays and may or may not make their first flights between this year and next.

Amazon said in its announcement that “the additional launches with SpaceX offer even more capacity to support our deployment schedule.” The company has said its Project Kuiper constellation will consist of 3,236 satellites, at least half of which must be in operation by summer 2026 to comply with its FCC license.

This article originally appeared on Engadget at https://www.engadget.com/amazon-is-swallowing-its-pride-to-ensure-its-internet-satellites-get-to-orbit-on-time-222906412.html?src=rss

An SEC filing has revealed more details on a data breach affecting 23andMe users that was disclosed earlier this fall. The company says its investigation found hackers were able to access the accounts of roughly 0.1 percent of its userbase, or about 14,000 of its 14 million total customers, TechCrunch notes. On top of that, the attackers were able to exploit 23andMe’s opt-in DNA Relatives (DNAR) feature, which matches users with their genetic relatives, to access information about millions of other users. A 23andMe spokesperson told Engadget that hackers accessed the DNAR profiles of roughly 5.5 million customers this way, plus Family Tree profile information from 1.4 million DNA Relative participants.

DNAR Profiles contain sensitive details including self-reported information like display names and locations, as well as shared DNA percentages for DNA Relatives matches, family names, predicted relationships and ancestry reports. Family Tree profiles contain display names and relationship labels, plus other information that a user may choose to add, including birth year and location. When the breach was first revealed in October, the company said its investigation “found that no genetic testing results have been leaked.” 

According to the new filing, the data “generally included ancestry information, and, for a subset of those accounts, health-related information based upon the user’s genetics.” All of this was obtained through a credential-stuffing attack, in which hackers used login information from other, previously compromised websites to access those users’ accounts on other sites. In doing this, the filing says, “the threat actor also accessed a significant number of files containing profile information about other users’ ancestry that such users chose to share when opting in to 23andMe’s DNA Relatives feature and posted certain information online.”

Following the discovery of the breach, 23andMe instructed affected users to change their passwords and later rolled out two-factor authentication for all of its customers. In another update on Friday, 23andMe said it had completed the investigation and is notifying everyone who was affected. The company also wrote in the filing that it “believes that the threat actor activity is contained,” and is working to have the publicly-posted information taken down.

Update, December 2 2023, 7:03PM ET: This story has been updated to include information provided by a 23andMe spokesperson on the scope of the breach and the number of DNA Relative participants affected.

This article originally appeared on Engadget at https://www.engadget.com/23andme-hackers-accessed-ancestry-information-from-thousands-of-customers-and-their-dna-relatives-205758731.html?src=rss