Open

Category Archives: Tech

Tamara News  /  Blog  /  Tech
02 Sep 2026
Amazon fulfillment center exterior linked to the FTC Amazon advertiser overcharging case

The FTC Says Amazon Secretly Overcharged Advertisers by $20 Billion — Here’s How

Categories Business, Tech

The FTC Amazon advertiser overcharging case, filed Monday in Seattle federal court, accuses the company of secretly manipulating its ad auction system. Regulators say the scheme squeezed more than $20 billion out of advertisers since 2019. The Federal Trade Commission and 22 state attorneys general brought the suit together.

The complaint says Amazon misled roughly 1.2 million advertising customers about how pricing and terms worked for its sponsored listings. More than 500,000 small and medium businesses are among those affected. Sponsored listings are the ads that appear at the top of search results across Amazon’s marketplace.

What the FTC Amazon advertiser overcharging complaint alleges

According to the FTC, Amazon changed its ad auction rules in 2019 in a way that let it quietly raise the effective price advertisers paid. The company did not disclose the change, regulators allege. CNBC reports the complaint centers on sponsored product ads, brand ads, and display ads that run alongside search results. The hidden surcharges, the FTC argues, ultimately pushed up prices that consumers paid too.

Amazon fulfillment center exterior linked to the FTC Amazon advertiser overcharging case

Amazon JFK8 fulfillment center photo: Tdorante10, CC BY-SA 4.0, via Wikimedia Commons.

How Amazon is responding

Amazon calls the lawsuit “misguided.” The company says it fundamentally misunderstands how its advertising auctions work and presents no evidence that shoppers were harmed. Amazon’s auction system prioritizes ad relevance to a shopper’s search terms before bid size, the company says, a design meant to keep prices lower for advertisers overall. Amazon also cites figures showing its auction changes saved advertisers roughly $8 billion between 2021 and 2025.

Amazon logistics warehouse building tied to the FTC Amazon advertiser overcharging lawsuit

Amazon JFK8 logistics park photo: Tdorante10, CC BY-SA 4.0, via Wikimedia Commons.

Why this case adds to Amazon’s legal exposure

The advertiser lawsuit lands on top of other pending FTC actions against Amazon. One separate case alleges the company enrolled consumers in Prime memberships without clear consent. Antitrust lawyers say a loss in the advertiser case could force changes to how Amazon structures ad pricing across its marketplace. A court order could require more transparency in how the auction sets prices. The case could also lead to civil penalties running into the billions of dollars if the FTC prevails.

How advertisers and sellers are reacting

Trade groups representing small online sellers say the case validates complaints they have raised for years about opaque ad pricing on Amazon’s platform. Some advertising agencies that manage Amazon campaigns for clients say they plan to review historical billing data to estimate their own potential exposure to the alleged surcharges. Amazon has not said whether it will offer any settlement fund to advertisers while the case proceeds through the courts.

State attorneys general have increasingly joined federal cases against large technology companies in recent years. They argue that coordinated action gives regulators more leverage than either level of government pursuing a case alone. The 22 states involved in this lawsuit span both major political parties. Legal observers say that detail signals broad, bipartisan concern about the underlying allegations rather than a partisan enforcement effort. Amazon’s outside counsel has signaled the company will contest both the federal and state claims together rather than seeking to settle with individual states separately.

What happens next in the Amazon case

Amazon is expected to file a motion to dismiss in the coming months. Discovery in a case this size typically stretches well over a year before trial. Advertisers who believe they were affected are watching closely. A favorable ruling for the FTC could open the door to private damages claims layered on top of any government penalty. For related coverage on major tech legal exposure, see our reporting on the Meta multistate settlement and the ongoing wave of tech layoffs.

This is not Amazon’s first brush with federal antitrust scrutiny. The company has faced separate government actions over its marketplace practices and subscription enrollment methods in recent years. Legal analysts say a pattern of parallel cases can add pressure on a company even when no single case is decisive. Courts and regulators sometimes reference findings from one matter when evaluating another. Amazon’s legal team has consistently argued that each case should be judged on its own facts rather than as part of a broader pattern.

Answering the obvious questions

How much did Amazon allegedly overcharge advertisers?
The FTC and state attorneys general allege Amazon extracted more than $20 billion in hidden surcharges from advertisers since 2019.

How many advertisers does the case cover?
The complaint says roughly 1.2 million advertising customers were affected, including more than 500,000 small and medium businesses.

What does Amazon say in response?
Amazon calls the lawsuit misguided and says its auction system saved advertisers billions of dollars rather than costing them money.

Who filed the lawsuit?
The Federal Trade Commission filed the case jointly with 22 state attorneys general in federal court in Seattle.

Could consumers be affected too?
The FTC argues the hidden ad surcharges fed into higher prices for shoppers. Amazon disputes that shoppers were harmed.

How long could the case take?
Cases of this size typically take well over a year to reach trial once motions and discovery are underway.

Consumer advocacy groups have welcomed the lawsuit, saying it validates years of complaints about opaque digital advertising markets generally, not just Amazon’s practices. Some economists who study online marketplaces say hidden auction mechanics are difficult for outside researchers to audit without regulatory subpoena power. That is part of why cases like this one often start with a government investigation rather than a private lawsuit.

Sources:

01 Sep 2026
Rocket launch of the kind that carried the Roman Space Telescope to orbit

NASA Just Launched a Telescope Aimed at Dark Energy

Categories Tech

NASA’s newest observatory is on its way. The Roman Space Telescope lifted off at 7:26 a.m. EDT on Sunday 30 August 2026, riding a SpaceX Falcon Heavy from Launch Complex 39A at Kennedy Space Center in Florida. The spacecraft reached orbit about ten minutes later.

Named for Nancy Grace Roman, the astronomer who shaped NASA’s early space astronomy programme, the bus-sized observatory is built to look for things other telescopes cannot easily find. NASA has described it as a discovery machine.

Deep field of stars and galaxies the Roman Space Telescope will survey
The observatory will spend five years surveying faint and distant objects.

How the Roman Space Telescope reached orbit

The Falcon Heavy is the heaviest-lift rocket SpaceX currently flies. Its 27 Merlin engines generated more than five million pounds of thrust at liftoff, pushing the stack through the atmosphere and away from Florida’s spaceport.

The rocket’s first stage consists of a centre core and two side boosters. The second stage, mounted atop the centre core, carried the observatory itself. The side boosters began their return shortly after separation, following the pattern SpaceX has flown on previous Falcon Heavy missions.

Reaching orbit ten minutes after launch is the first of several milestones. It confirms the ascent worked. It does not confirm the observatory is healthy, which takes considerably longer to establish.

The Roman telescope’s hundred-day cruise to L2

The telescope is not staying in Earth orbit. It is heading for the Sun-Earth L2 point, roughly 1.5 million kilometres from Earth, on a cruise NASA expects to take about 100 days.

L2 is a gravitational balance point that lets a spacecraft hold a stable position relative to the Earth and Sun while keeping the Sun, Earth and Moon on the same side. That geometry makes it easier to shield sensitive instruments from heat and light. The James Webb Space Telescope operates from the same region.

Arrival is not the end of the sequence. Commissioning follows, and instruments have to be cooled, aligned and calibrated before the survey can begin in earnest.

The questions the Roman Space Telescope was built to answer

Over the next five years the telescope’s observations target two of the largest open problems in cosmology.

The first is dark energy, the effect driving the accelerating expansion of the universe. Nobody has explained it. Measuring how that expansion has changed over cosmic time is one route to narrowing the possibilities.

The second is dark matter, which appears to supply much of the structure of the cosmos while remaining invisible to direct observation. Mapping how matter is distributed across large volumes of sky helps constrain what dark matter can be.

Beyond that, the instrument is designed to detect faint objects that are otherwise hard to pick out. That capability has applications well outside cosmology, including the search for planets around other stars.

Why a wide survey is a different kind of instrument

Telescopes trade breadth against depth. An instrument tuned to stare at one target for a long time produces exquisite detail about that target and tells you nothing about the rest of the sky. A survey instrument does the opposite: it covers ground, repeatedly, and finds the objects worth staring at later.

Roman sits in the second camp. NASA built it to sweep across large areas and pick out faint objects that would otherwise go unrecorded. That design choice explains the science goals. Dark energy and dark matter are statistical problems, and you cannot solve a statistical problem from a handful of deep exposures.

It also explains why the results will arrive as catalogues rather than as single spectacular pictures. The value of a survey accumulates. Individual frames matter less than what the whole dataset shows once it is assembled and compared against models.

What comes after Roman’s commissioning

Expect a gap before the first images. A hundred-day cruise, followed by commissioning, puts early science well into 2027 on the current schedule. NASA typically releases a first-light image once instruments are performing to specification.

The wider context is a busy period for large space science missions flown on commercial rockets. This launch is another data point in that shift, and the economics of it are part of why SpaceX’s valuation has drawn so much attention this year.

The computing side matters too. Wide surveys generate enormous data volumes, and processing them at speed depends on hardware supply chains that are under real strain, as our report on the AI chip interconnect bottleneck set out. Astronomy is now downstream of the same constraints as everything else in computing, a theme running through our reporting on how large institutions are buying compute.

Reader questions about the Roman Space Telescope

When did the telescope launch?

At 7:26 a.m. EDT on 30 August 2026, aboard a SpaceX Falcon Heavy from Launch Complex 39A at Kennedy Space Center.

Where is it going?

To the Sun-Earth L2 point, about 1.5 million kilometres from Earth, on a cruise expected to take roughly 100 days.

How long is the mission?

NASA has described a five-year period of observations following commissioning.

Who was Nancy Grace Roman?

An astronomer whose work shaped NASA’s early space astronomy programme. The observatory is named after her.

Is this a replacement for Hubble or Webb?

No. It is a separate observatory with its own survey mission, operating from the same general region of space as Webb.

When will we see the first images?

Not immediately. The cruise and commissioning phases come first, which puts early results well after arrival at L2.

Sources

01 Sep 2026
EU AI Act transparency

One Month Into the EU’s New AI Rules, Here’s Who’s Actually Complying

Categories Tech

The European Union’s AI rulebook has had a month to bite. EU AI Act transparency obligations took effect August 2. The European Commission’s AI Office and national regulators began enforcing rules that require AI systems to disclose their own involvement. Four areas are covered: direct interaction with people, AI-generated content, emotion recognition and biometric categorization, and deepfakes or AI-generated text on public-interest matters.

In practice, that means chatbots must tell users they are talking to a machine. Deepfakes need visible labels. AI-generated or altered content must carry machine-readable marks so platforms and researchers can detect it. Companies that don’t comply face fines up to €15 million or 3% of global annual turnover, whichever is larger. National market surveillance authorities, the European AI Office and the European Data Protection Supervisor enforce the penalty structure in parallel.

EU AI Act transparency

What EU AI Act transparency actually requires day to day

The rule’s four categories cover most consumer-facing AI use cases already in wide circulation. A customer-service chatbot on a European retail site now needs a clear disclosure at first contact. A marketing image generated or substantially altered by AI needs a machine-readable mark, not just a small watermark a user might miss. Emotion-recognition and biometric categorization systems, used in some retail and workplace settings, now require explicit disclosure to the people being monitored.

The AI Office has also published a voluntary Code of Practice on Transparency of AI-Generated Content. Several major AI providers have already signed on. That gives companies a template for compliance, rather than requiring each to interpret the regulation from scratch.

How this compares to the US approach to AI regulation

The EU’s transparency-first approach contrasts with Washington’s posture. The Commerce Department has signaled new rules are coming for AI chips and semiconductors, not for AI-generated content disclosure. That US regulatory track targets hardware supply chains. Brussels is regulating the output layer instead — what users see and interact with, rather than what powers it underneath.

China has taken a third approach. It recently fined AI companion apps for what regulators called inappropriately close engagement with users, a move covered in our reporting on Beijing’s AI companion crackdown. Three major regulatory blocs are now pursuing three distinct enforcement philosophies for the same underlying technology.

What happens next for companies still catching up

National regulators have signaled a phased enforcement approach. They are prioritizing the largest platforms and most visible violations first, rather than pursuing every noncompliant chatbot at once. Legal advisers tracking the rollout expect the first public fines to land in the coming months. Those are likely to target companies that ignored disclosure requirements entirely, rather than those that made good-faith but imperfect attempts at compliance.

Smaller companies building on top of major AI providers face a practical question. Does their chosen model provider’s compliance cover their own product, or do they need separate disclosure measures? Legal guidance from law firms including Cooley and Stibbe generally advises deployers not to assume upstream compliance protects them. The obligation applies at the point of user interaction, not just at the model level.

What compliance looks like for a small business

A small e-commerce site running an AI chatbot for customer support does not need a legal department to comply. It does need a few concrete changes. Those include a visible statement at the start of a chat session that the user is talking to an AI system, a process for labeling AI-generated product images or marketing copy, and a documented record of which disclosure measures are in place in case a regulator asks. Firms building on top of major providers’ APIs generally still carry this obligation themselves. The rule targets the point of interaction with the end user, not the underlying model.

Industry advisers say the most common compliance gap so far is not malicious evasion but simple oversight. Companies adopted AI tools for internal efficiency and only later realized customer-facing outputs also fall under the transparency rule. The Code of Practice published by the AI Office is designed to close that gap. It gives smaller companies language and formatting they can adopt directly, rather than drafting disclosure text from scratch.

Companies operating only outside the EU are not automatically exempt either. The rule applies based on where users are located, not where the company is headquartered. A US or Asia-based platform with European customers still needs to meet the same disclosure standard for the portion of its user base interacting from within the bloc. Legal teams at global platforms are treating that geographic split as a genuine engineering requirement, not just a policy footnote.

FAQ

When did the EU AI Act transparency rules take effect?

August 2, 2026.

What four areas do the rules cover?

Direct interaction with individuals, AI-generated content, emotion recognition and biometric categorization, and deepfakes or AI-generated public-interest content.

What are the penalties for noncompliance?

Fines up to €15 million or 3% of global annual turnover, whichever is greater.

Who enforces the transparency rules?

National market surveillance authorities, the European AI Office, and the European Data Protection Supervisor.

Is there a compliance template companies can follow?

Yes, the AI Office published a voluntary Code of Practice on Transparency of AI-Generated Content that several major providers have signed.

Does using a compliant AI model automatically make my product compliant?

Not necessarily. Legal advisers say the disclosure obligation applies at the point of user interaction, so deployers generally need their own compliance measures.

01 Sep 2026
European Commission headquarters, where the ChatGPT DSA designation was announced

ChatGPT Just Landed on the EU’s Strictest Regulatory List

Categories Tech

The European Commission has placed ChatGPT in the top tier of its online platform rulebook. The ChatGPT DSA designation, announced on 31 August 2026, classifies OpenAI’s chatbot search function as a Very Large Online Search Engine under the Digital Services Act. Reddit and Roblox were designated Very Large Online Platforms on the same day. All three now face the strictest obligations the law contains.

Twenty-eight services now carry that status in Europe. Until this week, the list read mostly as a roll call of social networks, marketplaces and app stores. A general-purpose chatbot and a children’s gaming platform did not sit on it.

Phone screen with apps affected by the ChatGPT DSA designation
ChatGPT, Reddit and Roblox now sit in the same DSA tier as Europe’s largest social platforms.

What the ChatGPT DSA designation actually means

The Digital Services Act sets a single numerical trigger. A service that averages at least 45 million monthly users inside the European Union crosses into the highest compliance tier. Platforms report those figures themselves, and the Commission then verifies and acts on them.

ChatGPT’s search function averaged roughly 159 million monthly active users in the bloc over the six months ending March 2026, according to the figures the Commission relied on. Reddit reported 57.2 million. Roblox reported around 48 million. Each number sits above the threshold, so each triggered a designation.

The split in labels matters. ChatGPT was designated as a search engine rather than a platform, because the Commission assessed its search feature. Reddit and Roblox were designated as platforms. The obligations overlap heavily, but the framing tells you which part of each product Brussels is looking at.

Why Reddit and Roblox met the DSA threshold too

Reddit’s European audience has grown steadily as its content surfaced more often in general web search. Roblox has a different profile: a very young user base, user-generated worlds, and an in-game economy. Regulators have paid close attention to both characteristics.

Roblox’s designation is the one to watch. The DSA’s minor-protection provisions were written with exactly this kind of service in mind, and the company must now assess and document how its design choices affect children in Europe. That is a heavier lift than a policy update.

The obligations that follow a top-tier designation

Designated services have four months to comply. That clock runs to the end of November 2026. In that window each company must build out a compliance programme covering several areas.

  • Annual systemic risk assessments covering illegal content, the protection of minors, effects on users’ mental and physical wellbeing, fundamental rights, electoral processes and public security.
  • Risk mitigation measures that respond to whatever those assessments find.
  • Independent audits of compliance, paid for by the company and conducted by an outside firm.
  • Data access for regulators and vetted researchers, so outside parties can study how the service behaves.

The Commission supervises this tier directly rather than leaving it to national regulators. That is a deliberate design choice in the law, and it removes the forum-shopping that shaped earlier EU tech enforcement.

How the DSA designation sits alongside the EU AI Act

Two separate European laws now touch the same product. The DSA governs how a service handles content, risk and users. The AI Act governs how an AI system is built, documented and disclosed. OpenAI has to satisfy both, and the requirements do not map onto each other neatly.

We covered the transparency provisions that took effect this summer in our report on the EU AI Act transparency rules. Read alongside this designation, the direction is consistent: Europe is regulating AI products through general platform law as well as through AI-specific law.

The commercial context matters too. AI companies are raising and spending at a scale that makes European compliance costs a rounding error, as the DeepSeek funding round illustrated this week. Compliance is a schedule problem for these firms, not a budget problem.

Where the DSA compliance clock goes from here

Expect three things before December. First, published risk assessments, which tend to arrive close to the deadline. Second, product changes in Europe that arrive quietly, particularly around age assurance on Roblox. Third, a period in which the Commission decides whether what it received is adequate.

Enforcement, if it comes, follows later. The DSA allows fines of up to 6% of global annual turnover for serious breaches, and the Commission has shown it will open proceedings against designated services. Nothing suggests that is imminent here. The immediate story is a deadline, not a penalty.

For anyone building on these platforms, the practical effect is narrower than the headlines suggest. Users in Europe should see more reporting options, clearer explanations of recommendation systems, and stronger defaults for younger accounts. Users elsewhere may see some of the same changes, because splitting a product by region is expensive.

Questions readers are asking about the DSA designation

Does the ChatGPT DSA designation apply outside the EU?

No. The Digital Services Act applies to services offered in the European Union. Companies often roll changes out globally for simplicity, but the legal obligation stops at the EU border.

Why was ChatGPT designated a search engine rather than a platform?

The Commission assessed ChatGPT’s search function, which retrieves and presents information from the web. That fits the DSA’s definition of an online search engine, so the Very Large Online Search Engine label applied.

When exactly does compliance start?

Designated services get four months from designation. For these three, that runs to the end of November 2026.

What happens if a company misses the deadline?

The Commission can open formal proceedings, request information, and ultimately fine a service up to 6% of global annual turnover for serious infringements. In practice, enforcement usually begins with information requests.

How many services are now designated?

Twenty-eight services are designated as very large online platforms or search engines following this announcement.

Sources

More of our technology and regulation coverage: the Pentagon’s new generative AI platform.

01 Sep 2026
Pentagon GenAI.mil AI platform

The Pentagon Just Added ChatGPT and Grok to Its AI Toolkit

Categories Tech

The Pentagon widened its in-house AI lineup this week. On Monday, August 31, the Department of Defense activated OpenAI’s ChatGPT Mil and Starshield AI’s Grok for Government on its Pentagon GenAI.mil AI platform. Both join Google Gemini as available model families for military and civilian personnel. More than 1.7 million users already have accounts on the platform, which launched nine months ago.

ChatGPT Mil is accredited for controlled unclassified information at Impact Level 5. That is a Defense Department cloud-security standard for certain sensitive but unclassified work. A department official told Navy Times the tool is built for “document-heavy unclassified work” — planning, policy drafting, logistics and administration. Gemini remains the preferred option for search-style tasks.

Pentagon GenAI.mil AI platform

Why the Pentagon GenAI.mil AI platform now runs four models

Defense officials call multiple vendors a deliberate hedge. They don’t want to rely on one company for a capability the department now treats as core infrastructure. Adding Grok for Government alongside ChatGPT Mil “helps eliminate vendor lock-in and promote a broader American AI ecosystem,” per the department’s own framing, reported by DefenseScoop and TechCrunch.

Users can pick a model per task rather than being routed to one system. That flexibility mirrors what commercial AI buyers already do. Enterprises increasingly run more than one large language model provider side by side. The same pattern shows up in how Gulf businesses are adopting AI customer service tools.

What the expansion signals about defense AI spending

The GenAI.mil rollout arrives as Washington debates how tightly to regulate AI chips and infrastructure. Commerce Department officials have signaled new rules are coming for AI chip exports. That process directly affects which hardware the Pentagon and its contractors can build on. That regulatory shift sits upstream of decisions like this one. The models running on GenAI.mil ultimately depend on chip supply chains Washington is still rewriting.

The scale of adoption is notable on its own. Reaching 1.7 million unique users in nine months puts the Pentagon among the largest single enterprise AI deployments in the US government, ahead of most civilian agencies’ rollout pace.

What happens next for defense AI procurement

Expect further model additions rather than consolidation. Defense officials describe GenAI.mil as an open platform, built to onboard more frontier models as they clear security accreditation. It is not a fixed contract with one or two vendors. Congress and defense oversight bodies are likely to press for clarity on data handling. Impact Level 5 accreditation covers sensitive but still unclassified material, and lawmakers have previously raised concerns about commercial AI vendors’ access to government data at scale.

The broader AI chip bottleneck story is far from resolved, too. Industry reporting points to interconnect and networking hardware, not raw chip supply, as the binding constraint on scaling systems like this one. That dynamic is covered in detail here.

What oversight questions remain unanswered

Adding commercial vendors to a defense-wide AI platform raises questions beyond which model performs best. Data handling is the central one. Impact Level 5 accreditation governs how sensitive unclassified information can be processed. It does not, by itself, resolve whether prompts or usage patterns from GenAI.mil sessions could ever reach the underlying vendor for model improvement. Defense officials have not detailed vendor-specific data retention terms publicly. Congressional oversight committees have flagged similar questions in past reviews of large federal AI contracts.

There is a competitive angle too. The Pentagon’s endorsement gives OpenAI and xAI, alongside Google, a marquee government reference customer. Frontier AI vendors are competing hard for enterprise and government contracts right now. How rival vendors not yet on GenAI.mil respond will matter — whether they seek their own accreditation path or lobby for a more open procurement process. That response is likely to shape the platform’s vendor roster over the next accreditation cycle.

The rollout also lands against a backdrop of federal AI adoption moving faster than many expected two years ago. GenAI.mil’s growth to 1.7 million users in nine months outpaces most civilian agencies. Defense officials cite that scale as evidence the multi-vendor model is working as intended. Other federal agencies weighing their own AI rollouts are watching the Pentagon’s template closely as a possible model to copy. Some have already sent staff to observe GenAI.mil’s onboarding process firsthand. A few have asked for briefings on how the Pentagon structured its vendor accreditation timeline.

FAQ

What is GenAI.mil?

GenAI.mil is the Department of Defense’s enterprise AI portal, giving military and civilian personnel access to multiple frontier AI models for unclassified work.

Which AI models are now available on GenAI.mil?

Google Gemini, OpenAI’s ChatGPT Mil, and xAI/Starshield’s Grok for Government, as of August 31, 2026.

What security level is ChatGPT Mil accredited for?

Impact Level 5, a Defense Department standard covering certain sensitive but unclassified information.

How many people use GenAI.mil?

More than 1.7 million unique users, nine months after the platform launched.

Why did the Pentagon add multiple AI vendors instead of one?

Officials say it avoids reliance on a single vendor and supports a broader domestic AI ecosystem.

What is each model used for?

Gemini is positioned for search-style tasks, while ChatGPT Mil targets document-heavy work such as planning, policy and logistics.

1 … 20 21 22 23 24 … 26