Researchers Say BADBOX 3.0 Botnet Sent 40 Billion Fake Ad Bids a Day
Cheap streaming boxes carried hidden code that faked ad views and rented out owners' home internet, according to researchers who say they have disrupted the operation.
Cheap streaming boxes carried hidden code that faked ad views and rented out owners' home internet, according to researchers who say they have disrupted the operation.
The robotaxi company's federal filings follow a California ban on its freeway driving, and one of the software fixes is not ready yet.
A Chinese-linked hacker pointed automated programs at Asian universities, and one turned up a bug its developers had never heard of. Amazon says the hacker went on to gain control of a server.
Geneva talks on weapons that can attack without a person's final decision produced a two-year work plan rather than a ban, and campaigners say it is too weak.
The redesigned desktop page puts AI shortcuts under the search box and keeps recent conversations one click away, while normal searches still work with the Enter key.
A new report on a federal investigation into the chipmaker's books deepened a historic sell-off, and the company's market value has now fallen behind three rivals.
A report from Bloomberg says Google's video service has agreed in principle to a deal with thousands of schools that sued over harm to students, though key steps remain.
A financing secured by Nvidia chips collided with Elon Musk's plan for data centers in orbit, and creditors wanted a say in where the hardware goes.
The SpaceX-owned coding tool built its own model on a Chinese lab's technology, and congressional investigators say it has told them of a possible change.
The software vendor first blamed configuration problems for the multi-day disruption, then said an intruder had reached one of its servers. Some management tools are still offline.
The company has moved from offering the update on request to delivering it by itself, starting with machines that are not run by an IT department.
Signal is blocking screenshots and screen reading by phone AI assistants to keep chats private. It calls the fix incomplete and wants the makers of iOS, Android and Windows to give apps real opt-out controls.
Network-equipment maker F5 urged customers to apply special out-of-cycle hotfixes for serious BIG-IP bugs. The company offers no stopgap for the worst of them.
California is investigating whether Nvidia's chips or confidential data reached China, and whether investors were kept in the dark about that risk.
The world's biggest contract chipmaker is reportedly pausing some advanced orders from Chinese customers, one of the last paths to making cutting-edge AI processors.
Investigators say the group posed as corporate help desks and used automated voice calls to break into company networks during a two-year investigation.
Meta's Quest headsets now double as a home theater with a catalog of three-dimensional films, though only shoppers in one country can use it at first.
Newly public Commerce Department licensing records detail which Chinese companies were cleared to buy advanced Nvidia and AMD processors, and lawmakers from both parties are demanding answers.
Google is handing its AI assistant direct control of a desktop computer, though only paying subscribers with the right hardware and location can try it.
Google has picked January 2027 to extend its identity checks on Android app makers beyond the four countries where they began.
The task manager inherited from Wunderlist is being phased out, and people with personal Microsoft accounts will have to find another app.
The privacy authority says wearable cameras with AI helpers can film bystanders unnoticed and misadvise users, including during a staged medical emergency.
Administrators running web servers on PHP have new releases to install, plus stopgap settings for the flaw that could let outsiders run code without logging in.
Alphabet's Google has chosen a Jeff Bezos rocket for its plan to run artificial intelligence chips in orbit, though no price or satellite count was disclosed.
Rapid7 says an intruder moved from a Confluence server into cloud systems within hours, and Atlassian rates the flaw critical.
A gang called The Gentlemen has put the chipmaker on its extortion site with a ticking clock but no evidence, while Nvidia says its own networks look untouched.
The agency says its essential work continued, but the outage adds to a string of technology failures across the United States.
Security firms and a European computer emergency team are urging web operators to update quickly after researchers released working demonstrations against one of the bugs.
The new model targets coding and office work at a lower price than a sibling model, and an independent tester places it behind two rivals.
The cybersecurity company announced the all-stock acquisition alongside quarterly results and a higher full-year outlook, and its shares jumped in after-hours trading.
A bogus livestream tied to a Tesla event promised to double any crypto sent, and the exchange says an AI agent stepped in.
Hackers exploited a weakness at the firm that runs credit unions' web and app banking, and an Irish minister says the delay in telling them was unacceptable.
The new preview ships locked down out of the box, though Microsoft warns the stricter configuration could disrupt older software and tools.
A researcher found two bugs in an India-based platform used for collections by large lenders, and the exposed files included call recordings and legal papers.
Messages among Meta employees describe odd and possibly invasive behavior in the assistant, and outside testers say part of the problem persists.
Fifteen figures from security and AI policy want the Justice Department to examine the chipmaker's handling of warning signs, though the letter is a request, not a finding.
The automaker's first U.S.-made cells for a compact electric pickup are leaving the plant, and the Chinese licensing deal behind them is drawing criticism from Washington.
A new button in Google's AI Mode starts chats that stay out of account history and personalization, but they expire in 24 hours and some data collection continues.
Citibank's updated online agreement puts the risk of third-party AI tools on account holders, and several other large banks have moved the same way.
A congressional watchdog got the most-used AI platforms to draft workable plans for violent attacks, and the incoming Senate homeland security chair now plans hearings on AI safety.
Two tampered versions of a Microsoft library for AI agents carried a credential stealer, and the attacker uploaded a second after the first was pulled.
Internal presentations released in child-safety litigation describe a plan to grow Instagram through schools, and Meta calls the characterization wrong.
The AI music platform now builds short videos that follow a song's lyrics, a step beyond music as it faces lawsuits and some critical users.
The chipmaker disclosed the halt on an earnings call without giving a reason, even though Washington had approved sales of the processors last year.
Buyers still waiting for the high-tech bike face a longer delay while the startup works on a firmware fix for problems owners say drain batteries overnight.
The browser's new stable release lands with a long repair list, and Google gives no sign that attackers are using any of the flaws.
Google's threat analysts found an AI system that spots newly patched software flaws, writes attacks for them and launches them with no person involved.
A study by UNESCO and researchers at University College London found every AI model it pressured eventually backed away from correct climate science to side with the user.
The shoemaker says a new way to build its signature air cushion lets engineers dial in softness and support by area, and it also cuts how long prototypes take.
The chipmaker has cautioned customers about proposed legislation on remote use of its processors, as Commerce issues a separate rule meant to block Chinese access through foreign data centers.
Researchers planted fake credentials in hundreds of container images and watched how fast automated scanners found them. The fastest reaction took less time than a breath.
The repair site expected a smooth job based on recent Apple products. Instead its technicians broke a speaker and tore a connector while trying to get inside.
Microsoft used its Ignite keynote to recast Copilot as the place where work begins and started rolling out an always-on agent to early-access customers.
The weakness sits in a remote-access service, and flipping a switch in the admin screen does not protect older installations.
U.S. federal agencies were urged to patch within two weeks, and administrators are told to tighten a certificate setting that can widen the danger.
The startup, founded by an executive who once held a top post at Neuralink, plans to widen sales of its vision device in Europe and seek U.S. approval.
The Pentagon chose seven firms from 13 proposals to fire high-energy lasers at drones in Utah, with purchase orders possible for systems that perform.
The secretary of state told reporters in Asia that the failures hitting airlines, banks and software services have separate causes, as fear of a coordinated attack spread online.
With two House members from opposite parties, she urged leaders to move a bill that would let victims sue over sexually explicit AI images before time runs out.
A string of reversals on the council, set off by revelations about how police used the system, ended police use of the vendor's license plate readers.