Researcher plants 2m canaries into the AI supply chain. A privacy disaster surfaced.

Researchers planted 2 million digital tripwires in requests sent through a sprawling network of AI services to find out what happened to sensitive information entrusted to them. What they discovered was an ecosystem that leaked highly sensitive data like a sieve.

The exposure was found through an ingenious experiment designed to test if sensitive information can leak through the vast chain of widely used AI services and how that information is exploited when it does. Many of today’s AI apps use services such as OpenRouter or Vercel that act as brokers. These “AI routers,” sometimes also known as “AI gateways,” receive large numbers of prompts sent from AI app end users and route them to an available inference engine based on criteria such as the one charging the lowest price or being best suited for the given task. Instead of using Anthropic, OpenAI, or another big provider, the developer hires an AI router to find the most suitable model.

Security breaches, prompt theft, and privacy violations

To test for potential leakage, researcher HD Moore assembled a collection of AWS access keys, functioning credentials tied to empty AWS accounts, passwords for services such as SSH, and unique URLs, DNS names, and private git repository references. None of the secrets had ever been used before, and he controlled the infrastructure that received any requests using the credentials. He then seeded each secret into exactly one prompt sent to a single provider. In the event any of the secrets were leaked, these “canaries”—so called because they resemble the role of that bird when taken into coal mines—would allow Moore to identify each link in the chain that was responsible.

Read full article

Comments

Google’s latest search page redesign drops the search button, adds AI sidebar

Google’s flagship search experience is evolving into an AI experience with stunning speed. Even time-tested elements of search are getting AI makeovers, and the latest to get this treatment is the search homepage on Google.com. A redesign that’s been in testing for the last few weeks is now official, bringing easier access to AI chats and dropping the dedicated search button entirely.

Google search VP Robby Stein confirmed on X that Google is deploying the new desktop search page after a successful test. The page will now have a few more buttons under the search bar, all connected to AI features. There’s an image generation button, one for analyzing files, and one for brainstorming with Gemini. The “I’m Feeling Lucky” button remains, but the Google Search button that used to sit next to it is gone.

The search button has obviously been a part of Google’s search page forever. However, it wasn’t in a very convenient location in the modern interface. When you begin typing a query, the autocomplete suggestions cover that button. You have to click out of the text field to access the button, which is an extra step compared to simply mashing the Enter key to run your search.

Read full article

Comments

Meta reconsiders 2030 net-zero goal amid record emissions from AI data centers

Meta’s executives are rethinking the company’s original goal of achieving net-zero emissions by 2030. That revelation in Meta’s latest sustainability report comes as the company reported a record percentage increase in greenhouse gas emissions that exceeds any other Big Tech competitor racing to build new AI data centers.

The company’s 2026 sustainability report claims that “our long-term ambitions haven’t changed, but our operating environment has,” according to an introduction from Rachel Peterson, vice president of data centers at Meta, and Blair Swedeen, global head of net zero and sustainability at Meta. In the November 19, 2026, report, they attributed the changing conditions and the need for a rethink to how “demand for AI infrastructure has grown faster than almost anyone projected.”

“We are re-evaluating our timelines, sequencing, and interim milestones—including the 2030 timeframe for our net zero emissions goal and our near-term intensity targets—so our approach reflects current operating realities while keeping progress transparent year over year,” Peterson and Swedeen wrote in the report.

Read full article

Comments

This 65-Inch Samsung Mini-LED TV Is $200 Off Ahead of Black Friday

We may earn a commission from links on this page. Deal pricing and availability subject to change after time of publication.

Early Black Friday deals are popping up everywhere from Amazon to Costco. For those looking to upgrade an old but smaller TV to a newer model without splurging on an OLED, the 65-inch Samsung M80H Mini-LED TV is 29% off and down to a record-low $497.99 (originally $697.99). If you want to downsize (or go massive), the 55-inch, 75-inch, and 85-inch models are also discounted ahead of the holiday.

While it’s considered an entry-level 4K smart TV, the M80H is packed with Samsung Vision AI features, and like the rest of the brand’s TVs, it runs on the Tizen interface. A high 144Hz refresh rate, as well as VRR and  AMD FreeSync Premium, makes it a decent choice for gamers compared to the standard 60Hz of cheaper models.

It has AirPlay support and built-in Alexa, but lacks local dimming and may struggle in very bright, sunlit rooms, so keep that in mind if you do a lot of daytime viewing. That said, Filmmaker Mode delivers reliable color accuracy for standard dynamic range (SDR) content out of the box. At under $500, the 65-inch Samsung M80H Mini-LED TV is a solid option if you want a big screen and a high refresh rate on a budget.

Deals are selected by our commerce team

Trump Admin Still Won’t Let TP-Link Sell Routers In US

An anonymous reader quotes a report from Ars Technica: TP-Link, which may be the top provider of wireless routers for US homes and small businesses, isn’t allowed to sell its latest Wi-Fi routers in the US. The Trump administration imposed a sweeping ban on foreign-made consumer routers in March 2026, and TP-Link hasn’t been able to obtain an exemption. The rules prevent the import or sale of all new router models that are manufactured, assembled, designed, or developed outside the US. The Trump administration has control over which router makers receive special permission to import and sell new products through a waiver-review process involving the Federal Communications Commission, the Department of Defense, and the Department of Homeland Security. While some of its competitors have obtained exemptions to the ban, TP-Link appears to be having a hard time making its case to government officials. Its routers approved before the March 2026 ban remain available for sale in the US, but its newly released Wi-Fi 8 routers cannot be imported or sold. Earlier this year, TP-Link announced plans to begin assembling routers in Mississippi as early as 2027, “but convincing Trump officials that it is fully under US control is likely a difficult task,” reports Ars. “The government reviews of router makers’ requests for waivers, formally called conditional approvals, are shrouded in secrecy.”


Read more of this story at Slashdot.

Prisons get OK to jam contraband phones. CTIA says nearby phones will be jammed too.

The Federal Communications Commission yesterday voted to let state and local prisons use jamming technology to block contraband cellphones, despite warnings that jammers will also block lawful phone calls outside the prisons.

Wireless industry lobby group CTIA said a recent test it conducted with the Virginia Tech Applied Research Corporation and South Carolina Department of Corrections confirmed concerns that the jammers would block legitimate communications, including 911 calls. CTIA told the FCC in an October 1 filing:

The pilot test was limited in the number of micro-jammers deployed, the spectrum bands they were authorized to jam, and the power levels at which they were permitted to operate. Even with these constraints, the micro-jamming system blocked 911 calls, interfered with public safety and other communications, and caused interference that impacted wireless network infrastructure up to two and a half miles away. At the same time, the pilot results show substantial gaps in jamming the targeted “contraband” devices, with testing scenarios in outdoor inmate recreation areas showing failure rates—meaning “contraband” device call completion—in one-half to two-thirds of call attempts.

CTIA said the pilot test shows “that current jamming technology seems to be both over and underinclusive at once—it risks blocking lawful communications even as it fails to block a significant portion of the targeted contraband device communications.”

Read full article

Comments

Astronomers watched a star spiral to its death inside a giant companion

In the Andromeda Galaxy, a giant star about seven times as massive as the Sun spent nearly a decade slowly swallowing its much smaller companion. The process, known as common envelope evolution, was proposed in the 1970s to explain how some of the tightest pairs of stars in the Universe form, but no one had ever caught it in the act.

In a new Nature study, astronomers led by Viraj Karambelkar, an astrophysicist at Columbia University, finally witnessed the entire spectacle from the decades before it began to the eruption that ended it in late 2025. “It lasts for a few years, maybe a few decades,” Karambelkar said. “Stars evolve on time scales of millions of years. So, it’s really like a blip on the entire cosmic time scale.”

Shared envelopes

Some pairs of stars orbit so close to each other that they could not have formed at that distance, so their orbits must have shrunk. Common envelope evolution is the leading explanation for that shrinkage. When one star in a binary grows old, it expands into a giant, and its outer layers can engulf its companion. The two then share a single, loosely bound envelope of gas. As the engulfed star drags through it, it loses orbital energy to friction and spirals inward. Depending on how much of the envelope gets lost, the pair either ends up in a much tighter orbit or merges.

Read full article

Comments

NASA’s new Moon plan is taking shape, but some big questions remain

SpaceX’s successful recovery of its Starship rocket last weekend cleared a major hurdle on NASA’s path to the Moon, opening up possibilities for higher launch rates and an all-important, long-delayed orbital refueling demo next year.

For SpaceX’s purposes, Saturday’s mid-air rocket catch will help engineers learn how to rapidly reuse Starships. That should expedite SpaceX’s deployment of its newest generation of Starlink Internet satellites. NASA, meanwhile, believes a higher launch cadence will lead to faster learning as SpaceX develops a deep space version of Starship to ferry astronaut crews to the surface of the Moon.

Early next year, SpaceX plans to bring online a second Starship launch site at Kennedy Space Center, Florida. A second Starship factory is also under construction at Kennedy, joining SpaceX’s existing SpaceX production line at Starbase, Texas. With two launch sites, SpaceX will have the pieces in place to take the next leap with Starship.

Read full article

Comments

California Issues Subpoena to Nvidia Over Chip Sales to China

California Attorney General Rob Bonta has issued subpoenas to Nvidia and former CEO Jensen Huang regarding reports that they misled investors about advanced AI chips potentially reaching China. “The conduct of Nvidia and its former CEO raises bright red flags, including about the ways in which the company may have misled investors,” Bonta said in a press release. “My office is investigating whether Nvidia or Jensen Huang’s actions resulted in China obtaining advanced chips or confidential information — and if Nvidia misled investors and the public about this risk, in violation of the law.” The probe follows Huang’s resignation amid allegations involving a suspected Chinese spy and comes as the SEC reportedly conducts a similar investigation. The Hill reports: Bonta’s office said the information obtained through the subpoenas will help it to determine if Nvidia, which is based in Santa Clara, Calif., violated state law through its alleged actions. “Nvidia is in a messy, transnational situation but the underlying principle of our corporate securities law is clear: Companies cannot mislead their investors,” Bonta continued.

[…] Following the Huang scandal, the House Foreign Affairs Committee released documents Thursday showing the U.S. authorized the sale of nearly 900,000 Nvidia and AMD AI chips to Defense Department-designated Chinese military companies. An Nvidia spokesperson previously told The Hill, “The review and licensing process spanned more than half a year, with approvals subject to strict conditions and limited volumes.” “Despite receiving some licenses, we have since stopped shipping H200s to China,” the spokesperson added.


Read more of this story at Slashdot.