Author Archives: Anna Washenko

Amazon discovered a ‘high volume’ of CSAM in its AI training data but isn’t saying where it came from

Posted on by

The National Center for Missing and Exploited Children said it received more than 1 million reports of AI-related child sexual abuse material (CSAM) in 2025. The "vast majority" of that content was reported by Amazon, which found the material in its training data, according to an investigation by Bloomberg. In addition, Amazon said only that it obtained the inappropriate content from external sources used to train its AI services and claimed it could not provide any further details about where the CSAM came from. 

Amazon provided Engadget with the following statement to explain why it doesn’t have data that can provide any further action on what it found.

“When we set up this reporting channel in 2024, we informed NCMEC that we would not have sufficient information to create actionable reports, because of the third-party nature of the scanned data. The separate channel ensures that these reports would not dilute the efficacy of our other reporting channels. Because of how this data is sourced, we don't have the data that comprises an actionable report.”

"This is really an outlier," Fallon McNulty, executive director of NCMEC’s CyberTipline, told Bloomberg. The CyberTipline is where many types of US-based companies are legally required to report suspected CSAM. “Having such a high volume come in throughout the year begs a lot of questions about where the data is coming from, and what safeguards have been put in place.” She added that aside from Amazon, the AI-related reports the organization received from other companies last year included actionable data that it could pass along to law enforcement for next steps. Since Amazon isn’t disclosing sources, McNulty said its reports have proved “inactionable.”

Amazon provided Engadget with these additional details, which were first reported in Bloomberg:

“Amazon is committed to preventing CSAM across all of its businesses, and we are not aware of any instances of our models generating CSAM. In accordance with our commitments to responsible AI and the Generative AI Principles to Prevent Child Abuse, we take a deliberately cautious approach to scanning foundation model training data, including data from the public web, to identify and remove known CSAM and protect our customers. While our proactive safeguards cannot provide the same detail in NCMEC reports as consumer-facing tools, we stand by our commitment to responsible AI and will continue our work to prevent CSAM.”

The company also reiterated that “we intentionally use an over-inclusive threshold for scanning, which yields a high percentage of false positives” to explain the high volume of content the company reported.

Safety questions for minors have emerged as a critical concern for the artificial intelligence industry in recent months. CSAM has skyrocketed in NCMEC's records; compared with the more than 1 million AI-related reports the organization received last year, the 2024 total was 67,000 reports while 2023 only saw 4,700 reports. 

In addition to issues such as abusive content being used to train models, AI chatbots have also been implicated in several dangerous or tragic cases involving young users. OpenAI and Character.AI have both been sued after teenagers planned their suicides with those companies' platforms. Meta is also being sued for alleged failures to protect teen users from sexually explicit conversations with chatbots.

Update, January 30, 2026, 11:05AM ET: This story has been updated with several statements from Amazon.

This article originally appeared on Engadget at https://www.engadget.com/ai/amazon-discovered-a-high-volume-of-csam-in-its-ai-training-data-but-isnt-saying-where-it-came-from-224749228.html?src=rss

Publishers are blocking the Internet Archive for fear AI scrapers can use it as a workaround

Posted on by

The Internet Archive has often been a valuable resource for journalists, from it's finding records of deleted tweets or providing academic texts for background research. However, the advent of AI has created a new tension between the parties. A few major publications have begun blocking the nonprofit digital library's access to their content based on concerns that AI companies' bots are using the Internet Archive's collections to indirectly scrape their articles.

"A lot of these AI businesses are looking for readily available, structured databases of content," Robert Hahn, head of business affairs and licensing for The Guardian, told Nieman Lab. "The Internet Archive’s API would have been an obvious place to plug their own machines into and suck out the IP."

The New York Times took a similar step. "We are blocking the Internet Archive's bot from accessing the Times because the Wayback Machine provides unfettered access to Times content — including by AI companies — without authorization," a representative from the newspaper confirmed to Nieman Lab. Subscription-focused publication the Financial Times and social forum Reddit have also made moves to selectively block how the Internet Archive catalogs their material.

Many publishers have attempted to sue AI businesses for how they access content used to train large language models. To name a few just from the realm of journalism:

Other media outlets have sought financial deals before offering up their libraries as training material, although those arrangements seem to provide compensation to the publishing companies rather than the writers. And that's not even delving into the copyright and piracy issues also being fought against AI tools by other creative fields, from fiction writers to visual artists to musicians. The whole Nieman Lab story is well worth a read for anyone who has been following any of these creative industries’ responses to artificial intelligence.

This article originally appeared on Engadget at https://www.engadget.com/ai/publishers-are-blocking-the-internet-archive-for-fear-ai-scrapers-can-use-it-as-a-workaround-204001754.html?src=rss

Apple acquires Q.ai for a reported $2 billion

Posted on by

Apple has acquired Israel-based startup Q.ai, a move that could provide a much-needed boost to the tech giant's capabilities in artificial intelligence. Although Apple has not disclosed terms of the deal, sources told Financial Times that the arrangement is reportedly valued at nearly $2 billion. If that figure is accurate, the Q.ai acquisition marks Apple's second largest acquisition to date, followed by its purchase of Beats for $3 billion back in 2014.

Johny Srouji, Apple’s senior vice president of hardware technologies, said in a statement that Q.ai "is a remarkable company that is pioneering new and creative ways to use imaging and machine learning." Apple hasn't shared any specifics about how it plans to leverage the startup, but its past work indicates the possibility of Apple moving deeper into AI-powered wearables. "Patents filed by Q.ai show its technology being used in headphones or glasses, using 'facial skin micro movements' to communicate without talking," the Times reported. 

The startup's founding team, including CEO Aviad Maizels, will join Apple as part of the deal. This acquisition marks Maizels' second sale to Apple; he previously founded a three-dimensional hearing business called PrimeSense that Apple bought back in 2013.

For several months, many tech insiders have speculated that an acquisition might be Apple's best path forward to catching up in the AI race. In the company's Q3 earnings call in July 2025, CEO Tim Cook acknowledged that "We’re open to M&A that accelerates our roadmap." A deal like this one could eventually lead to Apple developing its own fully in-house AI chatbot rather than relying on a competitor like Google to power artificial intelligence in its Siri assistant.

This article originally appeared on Engadget at https://www.engadget.com/big-tech/apple-acquires-qai-for-a-reported-2-billion-190017949.html?src=rss

Halide co-founder joins Apple’s design team

Posted on by

Apple picked up an intriguing new member for its design team today in Sebastiaan de With, co-founder of the iPhone camera app Halide. He announced the move today on Threads, adding, "So excited to work with the very best team in the world on my favorite products."

The Halide app has caught our eye at Engadget at several points over the years. de With also is co-founder of Lux, which is Halide's parent company. The other Lux apps also have an emphasis on photography and videography, particularly on Apple devices. Prior to Halide, de With had done other work at Apple, collaborating on properties including iCloud, MobileMe and Find My apps. It's unclear if his exit will mean any notable changes for Halide, or for the Lux apps Kino, Spectre and Orion.

For a long time, Apple's design philosophy was personified by Jony Ive, who left the company in 2022. Since his departure, no single person has emerged as the face and voice of Apple's attitude toward design, which could be why recent moves such as Liquid Glass have been met with deeply divided reactions.

This article originally appeared on Engadget at https://www.engadget.com/big-tech/halide-co-founder-joins-apples-design-team-235023416.html?src=rss

Astronomers share new insights about the early universe via the Webb Space Telescope

Posted on by

Researchers using the James Webb Space Telescope have found a galaxy that is offering new data about the early stages of the universe's existence. The latest discovery shared by astronomers is about a bright galaxy dubbed MoM-z14. According to the team, this galaxy existed 280 million years after the Big Bang. 

The sounds like a long time, but in the context of the universe's estimated 13.8 billion years of existence, that's actually one of the closest examples astronomers have found to the Big Bang's occurrence. As a result, MoM-z14 can offer some insights and some surprises about what the early stages of the universe entailed.

"With Webb, we are able to see farther than humans ever have before, and it looks nothing like what we predicted, which is both challenging and exciting," lead author Rohan Naidu of Massachusetts Institute of Technology said. The findings about this galaxy were published in the Open Journal of Astrophysics.

The scientists were able to date MoM-z14 with Webb's Near-Infrared Spectrograph instrument, analyzing how light from the galaxy changed wavelengths as it traveled to reach the telescope. One of the initial questions sparked by this bright galaxy centers on the presence of nitrogen. Some early galaxies, including MoM-z14, have revealed higher nitrogen concentrations than scientists had projected was possible. Another topic of interest is about reionization, or the process of stars producing enough light or energy to permeate the dense hydrogen fog that existed in the early universe. 

“It’s an incredibly exciting time, with Webb revealing the early Universe like never before and showing us how much there still is to discover” said Pennsylvania State University graduate student and team member Yijia Li.

This article originally appeared on Engadget at https://www.engadget.com/science/space/astronomers-share-new-insights-about-the-early-universe-via-the-webb-space-telescope-213311848.html?src=rss

Mark Zuckerberg was initially opposed to parental controls for AI chatbots, according to legal filing

Posted on by

Meta has faced some serious questions about how it allows its underage users to interact with AI-powered chatbots. Most recently, internal communications obtained by the New Mexico Attorney General's Office revealed that although Meta CEO Mark Zuckerberg was opposed to the chatbots having "explicit" conversations with minors, he also rejected the idea of placing parental controls on the feature.

Reuters reported that in an exchange between two unnamed Meta employees, one wrote that we "pushed hard for parental controls to turn GenAI off – but GenAI leadership pushed back stating Mark decision.” In its statement to the publication, Meta accused the New Mexico Attorney General of "cherry picking documents to paint a flawed and inaccurate picture." New Mexico is suing Meta on charges that the company “failed to stem the tide of damaging sexual material and sexual propositions delivered to children;” the case is scheduled to go to trial in February.

Despite only being available for a brief time, Meta's chatbots have already accumulated quite a history of behavior that veers into offensive if not outright illegal. In April 2025, The Wall Street Journal released an investigation that found Meta's chatbots could engage in fantasy sex conversations with minors, or could be directed to mimic a minor and engage in sexual conversation. The report claimed that Zuckerberg had wanted looser guards implemented around Meta's chatbots, but a spokesperson denied that the company had overlooked protections for children and teens. 

Internal review documents revealed in August 2025 detailed several hypothetical situations of what chatbot behaviors would be permitted, and the lines between sensual and sexual seemed pretty hazy. The document also permitted the chatbots to argue racist concepts. At the time, a representative told Engadget that the offending passages were hypotheticals rather than actual policy, which doesn't really seem like much of an improvement, and that they were removed from the document. 

Despite the multiple instances of questionable use of the chatbots, Meta only decided to suspend teen accounts' access to them last week. The company said it is temporarily removing access while it develops the parental controls that Zuckerberg had allegedly rejected using.

"Parents have long been able to see if their teens have been chatting with AIs on Instagram, and in October we announced our plans to go further, building new tools to give parents more control over their teens’ experiences with AI characters," a representative from Meta said. "Last week we once again reinforced our commitment to delivering on our promise of parental controls for AI, pausing teen access to AI characters completely until the updated version is ready."

New Mexico filed this lawsuit against Meta in December 2023 on claims that the company's platforms failed to protect minors from harassment by adults. Internal documents revealed early on in that complaint revealed that 100,000 child users were harassed daily on Meta's services.

Update, January 27, 2025, 6:52PM ET: Added statement from Meta spokesperson.

Update, January 27, 2025, 6:15PM ET: Corrected misstated timeline of the New Mexico lawsuit, which was filed in December 2023, not December 2024.

This article originally appeared on Engadget at https://www.engadget.com/social-media/mark-zuckerberg-was-initially-opposed-to-parental-controls-for-ai-chatbots-according-to-legal-filing-230110214.html?src=rss

Adobe Photoshop upgrades its Firefly-powered generative-AI editing tools

Posted on by

Adobe Photoshop introduced some new features that are rolling out for creators today. As you'd expect from any service operator in this day and age, there's some AI involved. Adobe has improved the tools for Generative Fill, Generative Expand and Remove that are powered by its Firefly generative AI platform. Using these tools for image editing should now produce results in 2K resolution with fewer artifacts and increased detail all while delivering better matches for the provided prompts. The Reference Image option for Generative Fill has also been upgraded to deliver "geometry-aware results that better match the scene." 

 One of the other new updates is a beta version of Dynamic Text, which should allow simpler transformation of a text layer into a curved shape. Photoshop has also added new adjustment layers: Clarity, Dehaze and Grain. These allow non-destructive image editing on layers.

This article originally appeared on Engadget at https://www.engadget.com/apps/adobe-photoshop-upgrades-its-firefly-powered-generative-ai-editing-tools-213737915.html?src=rss

Sennheiser debuts new models of wired headphones and earbuds

Posted on by

Wireless audio has become the industry standard, but there are still options out there for people who prefer a wired connection. Two new choices joining the market come from Sennheiser, which has released the CX 80U wired earbuds and HD 400U wired over-ear headphones. These new takes on the company's previous models for wired listening have replaced the 3.5mm audio jack connector with a USB-C cable. Both sets support 24-bit, 96 kHz digital audio playback. They're compatible with a broad array of devices, including iOS, iPadOS, Android, ChromeOS, MacOS, Windows and SteamOS. 

Both of these items are priced at an entry level for a brand that might charge up to $500 for its higher-end headphones. The CX 80U earbuds cost $40 and the HD 400U headphones retail for $100. Both products are available starting today.

This article originally appeared on Engadget at https://www.engadget.com/audio/headphones/sennheiser-debuts-new-models-of-wired-headphones-and-earbuds-201245058.html?src=rss

Sonos introduces Amp Multi for complicated residential installs

Posted on by

Sonos has unveiled its first new product of 2026, the Amp Multi. This amplifier is a niche option for the owners of very large or complicated spaces, and it's being billed as professional grade option for residential audio installations. The Amp Multi has eight 125W outputs and four configurable zones, and each channel can support up to three Sonos Architectural speakers. In other words, that's a lot more audio than the average home needs. Even the Sonos Amp would probably be overkill for those of you living the apartment life.

The Amp Multi will be available "in the coming months," according to the company's press release, and there's no pricing information yet for the product listing on its website. But given the high-end customers this is targeting, expect the Amp Multi to cost a fair bit more than the $800 Sonos Amp.

Sonos has mostly been keeping its proverbial head down on the product side as it continues to address fallout from a bungled app redesign in 2024 that soured customers and put the company in dire straits. First there were layoffs, then the CEO left. Sonos' temporary chief exec, Tom Conrad, got the position permanently last summer. Once the business' position does stabilize at last, we will hopefully be hearing more positive updates from Sonos in the future.

This article originally appeared on Engadget at https://www.engadget.com/audio/sonos-introduces-amp-multi-for-complicated-residential-installs-191000421.html?src=rss

The Samsung Galaxy Z Trifold will cost $2,900 in the US

Posted on by

As promised, Samsung is bringing its Galaxy Z Trifold smartphone to shoppers in the US. The unique foldable goes on sale on January 30. 

The phone was announced along with its select markets' sale dates at the close of 2025. We didn't have price information for Samsung's domestic market at that time, and even though we were anticipating the Galaxy Z Trifold to cost a bundle, the actual cost is more absurd than we'd guessed. In reality, getting your hands on this slim triple-screen smartphone will run you $2,900. 

Putting aside the ludicrous expense, this smartphone is pretty dang cool. After his hands-on experience with the phone at CES 2026, Engadget senior writer Sam Rutherford had overall positive things to say about the device, despite it demanding some tradeoffs in bulk. To borrow his own phrase: "I kind of hate how much I like the Galaxy Z TriFold."

Maybe as this cutting edge design becomes easier to manufacture, prices will eventually drop into the realm of reality for the average consumer. But if you’re a person who would buy a separate smartphone and tablet, combining both into a single device might feel worth the price tag. For now, anyone willing to pay the premium in the US will receive the model that has 512GB of storage and only comes in a black color.

This article originally appeared on Engadget at https://www.engadget.com/mobile/smartphones/the-samsung-galaxy-z-trifold-will-cost-2900-in-the-us-140000013.html?src=rss