A huge debate is brewing in Silicon Valley over the proliferation of Chinese-made artificial intelligence tools, particularly “open-weight” AI systems that, by some measures, can compete with or even outperform some of the best US models. My WIRED colleague Hugo Lowell has written about the Trump administration’s internal debate on how to handle these Chinese models. Among AI companies in the Valley, the issue is proving even more divisive.

A top concern in both DC and the Valley relates to distillation, in which a less capable AI model is trained on the outputs of a more powerful one. In June, Anthropic accused the Chinese tech giant Alibaba of illicitly stealing its IP through distillation attacks. Then, earlier this week, the White House said that it believes the Beijing-based Moonshot AI had developed its Kimi K3 model by distilling Anthropic’s Fable 5 model.

Another big concern is how quickly China’s models are appearing and spreading. An open-weight AI model has its core components made public, so that it can be fine-tuned to suit a user’s needs. But they don’t have the kinds of guardrails on which Anthropic has been building its reputation. Yasir Atalan, deputy director and data fellow at the Center for International and Strategic Studies, points out that the main benefit of open-weight AI models is their speed of diffusion. They can spread especially easily “through Hugging Face, GitHub, cloud providers, local deployments, and third-party inference platforms,” he writes. If you’re Anthropic, and you’ve built a cult around safety and charge for access to your big expensive proprietary models, you have every reason to want to regulate this.

But some Silicon Valley startups—not the trillion-dollar ones like OpenAI and Anthropic—really don’t want the US government to put restrictions on these AI models. On Wednesday a group of over 200 startups called the Little Tech Association sent a letter to Michael Kratsios, science adviser to President Donald Trump, and US Commerce Secretary Howard Lutnick lobbying against an outright ban of open-weight AI models. The group, which includes famed startup incubator YCombinator, has argued in favor of certain safeguards but says that denying Americans access to AI models abroad would weaken US startups and create a monopoly among the AI giants.

Bill Gurley, the legendary tech investor and longtime partner at Benchmark Capital, has publicly argued in favor of letting “the free market work.” In a lengthy blog that offers a nice little history of open-source software, Gurley writes that open-weight models avoid lock-in, encourage true academic research, and are critical for capital-constrained startups.

“Every AI startup, every solo developer, every two-person team building a product on top of AI infrastructure depends on having access to good models at affordable prices,” Gurley says.

Chamath Palihapitiya, one of the All-In podcast hosts, wrote on X that “tricking the US Government to protect frontier labs’ business model by using a China boogeyman is a mistake…It is protecting the equity of 5,000 people who are investors in OAI and Ant at the sale of everyone else. This would be a terribly stupid decision.” His cohost and fellow VC Jason Calacanis piled on. “Daddy Trump protect us!!!!” he wrote on X, with an alarming number of crying-laughing emoji.

This stance from some of Silicon Valley’s most ruthless capitalists might at first seem counterintuitive. Why let a foreign adversary’s technology flourish in the US? It’s as if the US is up just 1-0 in the AI World Cup, a slightly uncomfortable lead, and the crowd is chanting for the opposing team to get a free kick.

On July 23, President Donald Trump stood in the White House and promised that electricity bills for American families would “actually come down,” even as power-hungry AI data centres spread across the grid. T

he vehicle was an expanded version of the Ratepayer Protection Pledge, a voluntary scheme first unveiled in March. What energy analysts noticed was mostly what it left out.

The pledge asks the companies building data centres, among them Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI, to fund or build the power infrastructure their facilities demand rather than passing the cost to existing ratepayers.

The administration had already signalled it would widen the scheme to the utilities, and the new version reaches, by the White House’s own count, nearly 200 additional signatories, including NextEra Energy, Duke Energy, rural cooperatives, and a group of Republican governors.

Trump claimed the commitment now covers roughly 80% of the power delivered to US homes and businesses, and that companies given the right to build their own plants could sell surplus energy back to the grid, nudging rates down. He offered no capacity targets, no timelines, and no measurable milestones.

The 💜 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!

That silence matters, because the strain on household bills is already well documented. AI data centres have pushed up power bills across parts of the industrial Midwest, and in the PJM Interconnection, the largest US grid operator, data centres accounted for $6.3bn of the $16.4bn in charges from the most recent capacity auction, roughly 38%, according to the grid’s independent market monitor.

“PJM is continuing to act like it’s business as usual,” Joseph Bowring, the monitor’s president, said of the shift. “You have to open your eyes and recognise that it is really a paradigm shift, and failing to do that imposes costs on other customers.”

The pledge is not binding, and that is the central objection. It carries no penalties and no compliance oversight, and a quirk of federal rules may stop signatories honouring it even where they want to.

Current interconnection tariffs socialise grid-upgrade costs across all customers, and as FirstEnergy argued in a 2026 filing to regulators, existing rules can prevent a company from covering its own infrastructure costs even if it chooses to.

Consumer advocates were blunt. Jesse Lee of the campaign group Climate Power called the pledge a “pinky promise,” and a Consumer Reports survey found that 75% of American adults lacked confidence that large developers would truly cover all their costs.

Researchers at the Brookings Institution added that federal statutes “cannot readily override” the state public utility commissions that actually set residential rates.

There is a further wrinkle. Some of the same companies signing the pledge have fought state-level rules that would force them to deliver on it, consumer groups say, which makes the voluntary version look less like generosity than like the softer of two options.

The White House has cast the plan as proof that the AI build-out can proceed without punishing households.

The forecasts are not reassuring either. The consultancy ICF has estimated that data centres could lift US electricity demand by 25% by 2030 and add as much as 40% to monthly bills over five years, and utilities are planning some $1.4 trillion in capital spending by the end of the decade to keep up.

Louisiana, for its part, projects $2.6bn in ratepayer savings over 15 years from its deal with Meta, a reminder that the local arithmetic can cut both ways.

Congress has taken its own run at the problem, with the House advancing a bill on data centre energy costs, though nothing on the books yet compels the hyperscalers to pay.

The one body that could give the pledge teeth is the Federal Energy Regulatory Commission, which has already begun to fast-track grid connections for large loads. In June, it ordered six regional grid operators to justify or reform how they charge those users, with a deadline in August.

Until those rules change, the pledge remains what its critics say it is: a promise made in a room, with no one obliged to keep it.

• GitHub to launch two-tier (public and private) bug bounty schemes form July 27 2026
• Change comes in response to rise in lower-quality, AI-generated reports
• VIP researchers will earn around 3-4x more per report

GitHub has confirmed plans to evolve its bug bounty program into a two-tier system, which will come into force for reports submitted on or after July 27, 2026.

Under the new scheme, the Microsoft-owned coding platform will add a lower-paying public program that’s available to the wider research community, under a higher-paying invitation-only program.

A lawsuit over Apple’s failure to deliver Apple Intelligence and Siri features is moving closer to a conclusion as a judge has provisionally approved the company’s settlement offer.

The class action suit filed in 2025 alleged that people had bought new iPhones expressly because of Apple’s promoting of Apple Intelligence features that it then did not deliver. Apple admitted in March 2025 that the new Siri features were delayed, but then in May 2025 offered a settlement.

Apple and the parties to the class action suit agreed to a $250 million settlement, but it had to be approved. Now in a filing in the US District Court, Northern District of California, judge Noel Wise has provisionally given approval.

The final approval and so the start of any pay out to buyers, though, is dependent on a final hearing which has now been announced for September 29, 2027. While Judge Wise has given overall approval, he ruled that the court will not decide on any amounts until that hearing.

It may be a quick final hearing, though, as the parties have already agreed to a payment of at least $25 per eligible device. It is possible, however, that this amount could rise to $95 per device if the number of claims filed is low.

There will be a way for users to formally apply to be part of the claim. But as MacRumors said on Friday yet the details of this have not been announced.

from the pathetic,-weak-men dept

Earlier this year FCC boss Brendan Carr launched a series of fake investigations into ABC because the network (1) hosted Democratic Texas Senate hopeful James Talarico on The View, (2) aired comedians who made fun of the president and his wife, and (3) occasionally engaged in journalism critical of Trump corruption.

The details of these investigations really don’t matter at this point. We’ve discussed how they’re legally incoherent, clearly violate the First Amendment, and involve Carr actively manufacturing false claims that ABC violated FCC rules it was long-exempt from. At other points it just involves Carr being incoherently racist on Trump’s behalf; such as claims that ABC didn’t eliminate its “DEI” requirements quickly enough.

Carr’s now dumping additional empty threats into the mix, claiming that his ongoing review of ABC’s local broadcast licenses will take into consideration the network’s refusal to air Trump’s recent prime time speech, which mostly just involved Trump spewing more false election fraud conspiracy theories surrounding his 2020 loss.

ABC and NBC wisely refused to air the speech live, knowing that helping to spread distrust in election integrity in real time would be the opposite of useful journalism. That made Trump mad, so he’s clearly urged Brendan Carr to levy some additional empty threats against ABC:

“I think when you have the President of the United States standing inside the White House delivering an important speech, I think that’s something that broadcasters should be carrying. And so, obviously, this is an issue,” Carr told reporters Wednesday. “There have been lots of concerns raised, including by members of Congress, about whether broadcasters and their decisions there comply with the public interest.”

Carr is somewhat vague here because he knows this is a bunch of bullshit.

Obviously it’s ABC’s First Amendment right to determine what it broadcasts and when. Carr has absolutely zero legal role in determining the scheduling lineup of a private company. Carr’s once again pretending that networks that refuse to pander to our mad idiot king will be subjected to FCC review of their public interest obligations affixed to ownership of public airwaves.

As we’ve mentioned countless times already, Carr doesn’t want any of this to actually head to court because he knows it’s an absolute loser on First Amendment grounds. The real goal remains to threaten U.S. media companies with costly and annoying legal headaches if they challenge Republicans or the unpopular president. It’s typical lazy autocrat stuff by weak men who are afraid of words.

When it comes to ABC, that’s still been embarrassingly effective. The company agreed to pay Trump a $16 million bribe in 2024 to settle a baseless lawsuit the company easily could have won. And more recently, ABC shows like The View have shied away from hosting any political candidates at all for risk of upsetting Trump.

Brendan Carr has openly stated in interviews he fancies himself a tough, pit bull enforcer; but as Trump’s health and political power wane, the threats will hold less and less weight. As a result you’ve already seen ABC execs start to show a backbone in their fight with Carr, openly pointing out how he colluded with local right wing broadcast affiliates to manufacture evidence suggesting ABC broke FCC rules (something I’m sure will play great in court).

Carr’s threats will become weaker and weaker until he’s ultimately booted from office by subsequent administrations, at which point he’ll fail upward to some mid-six figure job at a telecom or media think tank, where he’ll spend the rest of his life helping corporate America dismantle whatever’s left of competition, labor, and consumer protection standards.

One of the ironic things, for Carr, is that his authoritarian censorship and saber rattling often draws press and public attention away from all the other terrible things he’s doing, whether it’s destroying media consolidation limits, making life easier on robocallers, dismantling broadband consumer protection standards, or making it easier for giant shitty companies to run amok.

You’d like to think Carr ultimately faces some sort of meaningful accountability for being one of the most censorial, petty, captured, and authoritarian regulators in U.S. history, but I wouldn’t hold your breath.

Filed Under: brendan carr, censorship, fcc, first amendment, james talarico, media

Companies: abc, disney

Erling Haaland is not the only Norwegian built around power, precision and making life difficult for the opposition.

Sigberg Audio is making its U.S. audio show debut at Southwest Audio Fest 2026 with the Manta active loudspeaker system, Saranna active floorstander and 10D dual opposed subwoofer.

The Norwegian manufacturer takes a rather different approach from the growing number of wireless speakers promising an entire audio system inside two attractive cabinets. Sigberg focuses on Hypex nCore amplification, DSP crossovers, manual parametric EQ and controlled directivity designed to reduce unwanted interaction with the room.

There is no Wi-Fi streaming, Bluetooth or HDMI eARC. These are active high-end loudspeakers for listeners who want the amplifiers, crossovers and drivers engineered as one system, not another lifestyle product whose future depends upon an app remaining alive.

Sigberg Audio Manta

The Manta is a four way active system comprising two powered monitors and a pair of dedicated Sigberg bass modules.

Each monitor uses a 12-inch midbass driver and a 5.5-inch coaxial driver containing a 1-inch silk dome tweeter. Three channel Hypex nCore amplification provides 600 watts per speaker, while DSP handles crossover and equalization duties.

Sigberg describes the Manta as a dual cardioid design. Its vented enclosure is intended to reduce sound radiating toward the sides and rear of the cabinet, potentially limiting early reflections from nearby walls.

The company claims rearward and lateral attenuation of approximately 10dB to 25dB across portions of the 100Hz to 5kHz range. That will not magically remove the room, but it could make the Manta considerably easier to integrate than a conventional loudspeaker with broad rearward radiation.

Sigberg recommends positioning the Manta between 10 and 50cm (4-20 inches) from the front wall and at least 40cm (15 inches) from the side walls. The included stands tilt the speakers upward by four degrees.

Manta Key Specifications

• Design: 4-way active dual cardioid system
• Drivers: 12-inch midbass and 5.5-inch coaxial driver with 1-inch silk dome tweeter
• Amplification: Three channel Hypex nCore Class D with DSP
• Power: 600 watts per speaker
• Frequency Response (with subwoofers): 25Hz to 20kHz, ±2.5dB (anechoic)
• Recommended listening distance: 1.5 to 8 meters (5-26 feet)
• Max SPL: 122 dB per speaker
• EQ: Nine band manual parametric EQ
• Analog Inputs: RCA and balanced XLR
• Digital Inputs: Optical, coaxial S/PDIF and AES
• Dimensions (WxHxD): 360 x 600 x 350 mm (14 x 23.5 x 14 inches)
  • Total height on included stands: 107 cm (42 inches)
• Weight: 25 kg (55 lbs) per speaker

Sigberg Audio 10D

The Manta system being demonstrated in Dallas uses two Sigberg 10D subwoofers.

Each sealed cabinet contains two opposing 10 inch aluminum cone drivers powered by a DSP enabled Hypex nCore amplifier. The opposing drivers are designed to cancel mechanical forces and reduce cabinet vibration.

Advertisement. Scroll to continue reading.

Sigberg specifies a frequency response of 23Hz to 250Hz, with typical in room extension to approximately 18Hz. Claimed average output is approximately 117dB from 30Hz to 80Hz at one meter under CEA 2010 measurement conditions.

Each 10D weighs 27kg, or 59 pounds, and includes RCA, balanced XLR and speaker level inputs. It also offers nine band parametric EQ and connection presets for use with an A/V receiver, conventional preamplifier or Sigberg active speaker system.

Advertisement

Sigberg Subwoofer Key Specifications

	Sigberg Inkognito	Sigberg 10D
MSRP	$4,000 each	$5,000 each
Driver Type	12″ Scan-Speak Aluminium cone	Dual 10″ Aluminium cone
Enclosure Type	Sealed	Sealed
Frequency Response	23- 250Hz (±3dB); ~18Hz in-room	23- 250Hz (±3dB); ~18Hz in-room
Max SPL	113 dB	117 dB
Amplification	Hypex nCore, DSP enabled	Hypex nCore, DSP enabled
Crossover/EQ	9-band parametric	9-band parametric
Inputs	RCA, XLR balanced, High-level	RCA, XLR balanced, High-level
Dimensions (WxHxD)	650 x 540 x 163 mm (25.6 x 21 x 6.4 in)	360 x 370 x 410 mm (14 x 14.5 x 16 in)
Weight	22 kg (48 lbs)	27 kg (59 lbs)

Sigberg Audio Saranna

The Saranna packages much of the same technology inside a narrower full range floorstander.

A front mounted 8 inch coaxial driver combines the midbass and midrange unit with a horn loaded compression driver. Two rear mounted 8 inch woofers handle the lower frequencies inside a ported enclosure.

Each Saranna contains 600 watts of three channel Hypex nCore amplification, DSP crossovers and nine band parametric EQ.

Sigberg claims a frequency response of 28Hz to 20kHz, with typical in room extension down to 20Hz. Claimed maximum output is 116dB per speaker.

The rear woofers might suggest that the Saranna needs considerable breathing room, but Sigberg designed the system to benefit from boundary reinforcement. The company recommends placement relatively close to the front wall, generally between 15 and 50cm.

Advertisement

Saranna Key Specifications

• Design: 3-way directivity controlled floorstanding speaker with ported bass enclosure
• Front Driver: 8-inch coaxial midbass and midrange driver with compression driver
• Bass Drivers: Two rear mounted 8-inch woofers
• Amplification: Three channel Hypex nCore Class D with DSP
• Power: 600 watts per speaker
• Frequency Response: 28Hz to 20kHz, ±2.5dB
• Max SPL: 116 dB per speaker
• Recommended listening distance: 2-5 meters (6-16 feet)
• EQ: 9-band parametric EQ
• Analog Inputs: RCA and balanced XLR
• Digital Inputs: Optical, coaxial S/PDIF and AES
• Dimensions (WxHxD): 284 x 1100 x 340 mm (11.2 x 43.3 x 13.4 inches)
  • Total height on included base with feet: 114 cm (~45 inches). Both feet and base are optional to install.
• Weight: 44 kg (97 lbs) per speaker
• Finishes: Fade to Black or Shady White

Active Does Not Mean Wireless

The Manta and Saranna include their own amplification, DSP and active crossovers. Owners do not need separate power amplifiers, but they will still need a source component or preamplifier with volume control.

Both systems support RCA, balanced XLR, optical, coaxial S/PDIF and AES connections.

Their published specifications do not include Bluetooth, network streaming, HDMI eARC, automatic room correction or an internal phono stage.

That makes them less convenient than something from KEF, Dynaudio or Buchardt, but it also means the loudspeakers are not tied to a specific streaming platform. Owners can select or replace the front end without throwing away the amplification and loudspeaker system.

The manual nine band parametric EQ also provides considerable flexibility, although using it properly will require acoustic measurements, dealer assistance or some idea of what those frequency and Q controls actually do. Randomly moving sliders until everything looks exciting is not room correction.

Advertisement

Advertisement. Scroll to continue reading.

Pricing & Availability

Sigberg says its direct pricing includes worldwide shipping, applicable taxes, import fees and customs handling. The company also offers a 60-day home trial and 5-year warranty.

• Manta with two 10D subwoofers: $30,000/pair
• Manta with two Inkognito subwoofers: $28,000/pair
• Saranna: $27,000/pair
• 10D: $5,000 each
• Inkognito: $4,000 each

U.S. customers can purchase directly from Sigberg or contact DreamScapes A/V in Syracuse, New York.

Where Can You Hear Them?

The Sigberg Audio systems are being demonstrated inside the DreamScapes A/V Presidential Suite on the 10th floor of the Sheraton Dallas Hotel.

The Manta and 10D are confirmed for the room. The Saranna was announced for the show but was initially listed as pending customs clearance, so attendees should confirm its arrival before heading upstairs specifically to hear it.

Advertisement

Southwest Audio Fest 2026 runs from July 23 through July 25 at the Sheraton Dallas Hotel. Tickets cost $25 for one day or $40 for a multiday pass.

The Bottom Line

Sigberg Audio is not trying to build another attractive wireless speaker that promises convenience above everything else.

The Manta and Saranna combine purpose matched amplification, DSP crossovers, substantial output and controlled directivity in systems designed to work relatively close to the front wall. That could make them especially interesting for listeners who want full range performance but do not have an acoustically perfect listening room the size of a Norwegian football pitch.

The Manta is the more ambitious system, combining large active monitors with two dedicated bass modules. The Saranna delivers much of the same philosophy in a narrower floorstanding design without requiring external subwoofers.

Neither system is inexpensive. There is also no onboard streaming, HDMI eARC or automatic room correction, and anyone who changes amplifiers more often than Erling Haaland scores goals will probably struggle with the concept.

Advertisement

For listeners who care more about engineering, room integration and consistent system matching than collecting amplifiers, Sigberg Audio could be one of the most interesting manufacturers making its Dallas debut.

For more information: sigbergaudio.com

Related Reading:

Advertisement. Scroll to continue reading.

Advertisement