from the such-a-waste-of-time dept

On Wednesday of last week, I wrote a post about how the Trump administration had quietly given up defending its unconstitutional executive orders targeting law firms. The DOJ was dropping its appeals, the firms that fought had won, and the firms that capitulated—led by Paul Weiss and their nearly $1 billion in groveling pro bono commitments—were left holding a very expensive bag.

What I did not realize, because this administration launches new absurdities faster than any human being can reasonably track them, was that the day before I published that piece (but about the time I was writing it), the DOJ had already filed a motion to take back its voluntary dismissal. And then, by Friday, the DOJ had filed a full appellate brief seeking to overturn all four district court rulings that struck down the executive orders.

So, to recap the timeline here: On Monday, the DOJ told the DC Circuit it was voluntarily dropping its appeals. All four law firms agreed. Done deal. On Tuesday, the DOJ filed a motion to withdraw its own voluntary dismissal. On Wednesday, I published an article mocking the administration for giving up. On Friday, the DOJ filed a 97-page opening brief arguing that the executive orders were “well within the Presidential prerogative.”

My only defense for coming in a day late is that covering this administration in anything close to real time is effectively impossible.

Let’s start with the procedural absurdity before we get to the substance—because the procedural absurdity is really something.

The motion to withdraw the voluntary dismissal is a remarkable document, mostly for how little it says. The entire operative section is barely over a hundred words. After all parties had agreed to the dismissal, the DOJ simply asked to take it back, offering no explanation whatsoever. The law firms’ collective response, included in the filing itself, was about as polite as you’d expect:

“Plaintiffs-Appellees oppose the government’s unexplained request to withdraw yesterday’s voluntary dismissal, to which all parties had agreed. Under no circumstances should the government’s unexplained about-face provide a basis for an extension of its brief.”

“Unexplained.” That word does a lot of heavy lifting. The DOJ’s motion doesn’t even try to explain why it changed course. There’s no “upon further reflection” or “new developments have arisen.” Just: forget what we agreed to yesterday, the court hasn’t formally granted the dismissal yet, so we’d like to un-dismiss please.

As of this writing, the court hasn’t ruled on that motion. But the DOJ apparently decided not to wait around and went ahead and filed its full appellate brief on Friday anyway.

The opening paragraph of the DOJ’s appellate brief is genuinely one of the more audacious things I’ve read in a legal filing, and I say that as someone who reads a lot of legal filings:

Courts cannot tell the President what to say. Courts cannot tell the President what not to say. They cannot tell the President how to handle national security clearances. And they cannot interfere with Presidential directives instructing agencies to investigate racial discrimination that violates federal civil rights laws.

Let’s focus on those first two sentences, because they reveal something important about how the administration is framing this case—and how badly they’re getting the First Amendment backwards.

“Courts cannot tell the President what to say. Courts cannot tell the President what not to say.”

Well, sure. In the most general sense, that’s true. The president can stand at a podium and say whatever he wants. He can say mean things about law firms. He can call them names on social media. He can go on television and express his displeasure with their client choices. That’s all government speech, and it’s all fine.

But that’s… not what happened here. What happened here is that the president issued executive orders imposing concrete sanctions on specific law firms—revoking security clearances, directing the termination of government contracts, restricting access to federal buildings, banning the hiring of their employees—because those firms represented clients the president didn’t like and employed lawyers who had been involved in investigations the president found personally disagreeable.

The brief tries to frame the courts’ injunctions as an attempt to “silence” the president. But nobody is trying to silence the president. The president can talk about these law firms every day from now until the world ends. What the courts said—four separate times—is that the president cannot use the machinery of government to punish law firms for their constitutionally protected legal advocacy. There’s a rather fundamental difference between speech and sanctions, and pretending not to understand that difference is doing a lot of work in this brief.

This gets at something we talk about regularly here at Techdirt: the First Amendment is a restraint on government power. It prevents the government from using its authority to suppress or punish private speech. When the DOJ frames this as courts trying to control the president’s speech, they’ve got the vector of the First Amendment claim pointing in exactly the wrong direction. The law firms aren’t saying the president can’t talk. They’re saying the president can’t retaliate against them for their own protected speech and advocacy. Those are two wildly different things.

The brief actually cites NRA v. Vullo, the 2024 Supreme Court case that we’ve written about a few times. For those unfamiliar, that case involved New York’s former superintendent of financial services, who was accused of using her regulatory power to coerce financial institutions into cutting ties with the NRA because she disagreed with the NRA’s advocacy. The Supreme Court held—unanimously—that government officials using their regulatory authority to punish or suppress disfavored private speech can violate the First Amendment, even if the official frames their actions in terms of legitimate regulatory interests.

The DOJ cites Vullo in the context of arguing that the district courts went too far in enjoining “future actions” based on Section 1 of the executive orders, quoting the district court’s ruling in favor of one of the law firms (Jenner & Block) favorably:

Significantly, even the district court in Jenner recognized this. That court declined to “enjoin future actions taken pursuant to Section 1,” because “Section 1 does not direct any action.” JA2205–06. But “shorn of its enforcement mechanisms, Section 1 is nothing more than the Executive Branch ‘saying what it wishes.’” Id. (quoting Nat’l Rifle Ass’n of Am. v. Vullo, 602 U.S. 175, 187 (2024)). “Jenner has no more right to silence the Executive Branch than the Executive Branch has to silence Jenner.” Id. That is because Section 1 is “government speech.” Id. Despite Jenner’s repeated request to enjoin Section 1 in the abstract, the district court correctly recognized that “[n]either standing doctrine nor equity generally permits such judicial prophylaxis.” JA2207. Thus, “[w]hether best viewed as a shortcoming of standing, ripeness, or” the lack of any basis in equity, “the guesswork entailed in enjoining all future uses of the sentiments expressed in Section 1 would exceed the Court’s proper role.”

The problem is that Vullo actually undercuts their entire argument. The point of the Vullo framework is that when government speech is coupled with government action designed to punish disfavored private expression, the combination can be unconstitutional coercion. The administration wants to unbundle its speech from its sanctions and defend each in isolation—”Section 1 is just government speech.” That’s precisely the move Vullo says you can’t get away with.

Meanwhile, I have to call out that the same people who argued in the Murthy v. Missouri case that any government speech criticizing private companies constituted a de facto First Amendment violation are now arguing “well, this paragraph was just speech, not retaliatory, so leave it alone.”

The brief also contains a line that should make Paul Weiss and others in the capitulation crowd feel especially great about their choices:

In recognition of those problems, many law firms agreed to address their practices and commit to providing pro bono work in the public interest.

The brief then helpfully lists them in a footnote in case anyone forgot which capitulating law firms to shun:

Allen Overy Shearman Sterling, Cadwalader, Kirkland & Ellis, Latham & Watkins, Milbank, Paul Weiss, Simpson Thacher, Skadden, and Wilkie Farr & Gallagher.

The DOJ is literally using the capitulation of those firms as evidence that the executive orders were reasonable and justified. “See? These firms agreed with us!” The firms that folded bought themselves a supporting role in the government’s brief arguing for the constitutionality of retaliating against law firms. Congratulations! Great job lawyering, guys.

Meanwhile, the four firms named in the brief who fought—Perkins Coie, Jenner & Block, WilmerHale, and Susman Godfrey—are named as parties who “instead filed suit.” See? Capitulating is the only proper move to this DOJ. Standing up for your own constitutional rights deserves punishment.

The heart of the filing is that opening framing. “Courts cannot tell the President what to say.” And the response to that is simple: nobody’s trying to. What courts can do—what they’re required to do under the First Amendment—is tell the president he cannot use executive power to punish private parties for their constitutionally protected advocacy. The fact that the DOJ appears unable or unwilling to understand this distinction tells you everything about the strength of their legal position.

As I noted last week, the administration’s decision to initially drop these appeals suggested that even a DOJ willing to argue almost anything looked at these cases and concluded it couldn’t win. The un-dropping and subsequent brief don’t change that calculus. While the DOJ offered no explanation for its reversal, the timing strongly suggests someone higher up didn’t like the press coverage of them folding and decided the political upside of continuing to threaten the legal profession outweighed the legal downside of losing again. Which, if you think about it, proves exactly what the law firms argued from the start: this was always about intimidation, never about law.

The firms that folded will keep being cited in government briefs as proof that the intimidation campaign was justified.

That’s the tax you pay for cowardice: your surrender becomes someone else’s evidence.

Filed Under: 1st amendment, doj, free speech, law firms, vullo

Companies: jenner and block, kirkland & ellis, latham and watkins, milbank, paul weiss, perkins coie, skadden, susman godfrey, wilkie farr, wilmerhale

Spend about ten minutes walking the floor at CanJam NYC 2026 and something becomes painfully obvious: wired IEMs are not quietly fading away into the Bluetooth sunset. If anything, they’re staging a very loud comeback. That only their users can hear but you understand where I’m going with this.

We counted roughly three dozen IEM brands showing new models this year and not just niche boutique outfits either. Meze Audio, Campfire Audio, Noble, 64 Audio, Astell&Kern, FiiO, Melody, Final DUNU, and a handful of smaller boutique builders were all pushing new wired designs that ranged from “somewhat affordable” to “this probably requires a second mortgage.”

This is happening at the exact same moment society has fully embraced streaming and wireless convenience. Walk down any street, get on the subway, or sit in a coffee shop and you’ll see the same thing: people glued to their phones with AirPods, Sony, or Bose wireless earbuds jammed in their ears while Spotify algorithmically decides what they should listen to next. Convenience won that war years ago.

But here’s the part that makes the CanJam floor so interesting. Despite the dominance of wireless listening, there appears to be a growing group of listeners who still care about sound quality enough to deal with the dreaded cable. And unlike full size headphones, wired IEMs solve a problem that a lot of audiophile gear doesn’t: they’re small, portable, and actually practical to carry around.

Pair them with one of the many dongle DACs or portable DAC/amps that have flooded the market over the past few years from companies like FiiO, Questyle, iFi, and Astell&Kern, and suddenly you have something that delivers far better sound than most wireless earbuds while still fitting in your pocket.

So yes, the rest of the world may be happily living in a wireless ecosystem fueled by streaming and convenience. But if the crowds around the IEM tables at CanJam NYC 2026 were any indication, there are still plenty of people willing to deal with a cable if it means their music actually sounds better.

And based on the sheer number of new wired IEMs launching right now, manufacturers seem to think that number is growing — not shrinking.

Which raises an uncomfortable question for the wireless everything crowd:

What if wired IEMs never actually went away? They just waited for everyone to remember what better sound actually feels like.

A Wall of IEMs: More Choice Than Ever for Listeners

I’ve never been the world’s biggest IEM fan. The whole “shove this into your ear canal and enjoy” concept never really worked for me. Some people swear by it. I usually spend the first five minutes wondering why my ears feel like they’re being fitted for dental molds.

The over-ear cable loop was always the least offensive part of the experience. It kept things relatively secure and avoided that lovely moment when someone brushes past you on a train platform or airport concourse and suddenly your headphones are being violently introduced to Newton’s laws of motion. Anyone who owned early fixed-cable IEMs knows the feeling: one snag, one sharp tug, and that cable is done.

So a sincere thank you to whoever finally realized detachable cables were not a luxury feature but basic survival equipment. The pro audio world figured that out decades ago. You can’t exactly be onstage in front of 90,000 people and have your monitor connection disappear because someone stepped on a cable. Robust connectors and replaceable cables were inevitable — and long overdue.

Yes, wireless will replace most of this eventually. Convenience tends to win those battles. But what makes IEMs fascinating in 2026 isn’t the cable debate — it’s the sheer level of innovation packed into something smaller than a pinky ring. Balanced armatures, planar drivers, electrostatic elements, hybrid designs mixing multiple technologies, and configurations that stack five, eight, ten drivers or more inside a headshell that looks like it belongs on a piece of jewelry.

It’s absurd engineering in miniature. And judging by what we saw at CanJam NYC 2026, the people building these things are just getting started.

One of the reasons we brought columnist Aaron Sigal back to cover wired and wireless IEMs is simple: the traffic is there. The demand is there. Our recent reviews of the Campfire Audio Andromeda 10, DUNU DN142, Apos x Community Rock Lobster, SIVGA Nightingale Pro, and Beyerdynamic DT 7x IE Series have all pulled in highly focused readership. Not casual drive by traffic either. The kind of readers who actually care about what driver configuration is inside the shell and whether the tuning leans reference or warm.

Is a lot of that coming from the Head-Fi crowd? Maybe. They can circle the wagons and argue endlessly about tips, cables, and impedance curves like it’s a graduate seminar in ear canal acoustics. But the interest is real, and the traffic numbers back that up.

Walking the floor at CanJam NYC 2026 made that even harder to ignore. There were so many tables dedicated to wired IEMs that it almost discouraged me from trying to cover them all. Full size headphones are still where my personal interest leans, and frankly that’s where a lot of our readers tend to focus as well. But the reality on the show floor told a different story.

The IEM tables were packed. Constantly.

Yes, the entire show was a sea of people moving from booth to booth, but the crowds hovering over those tiny display trays full of in ear monitors never really thinned out. People waiting for a chance to listen. Swapping tips. Plugging into portable DACs. Comparing notes.

Based on what we saw, it’s hard to imagine that the companies building wired IEMs didn’t have a very good weekend in New York. There’s just no way those tables were that busy if nobody was buying.

So here’s the question for readers.

Do you actually use wired IEMs? And if you do, why?

Is it about sound quality? Portability? Isolation on planes and trains? Or are you pairing them with a dongle DAC or portable player because you simply refuse to let Bluetooth compression have the final say in how your music sounds?

And let’s address the elephant sitting in the display case: price. The number of wired IEMs that now cost well into the thousands of dollars is…kind of insane. Universal or custom, it doesn’t seem to matter anymore. Some of these models cost as much as a very good stereo system or a pair of flagship headphones.

Does that discourage you? Or do you see them as the most practical way to get reference level sound in a portable format?

We’re genuinely curious where people land on this.

Related Reading:

Data brokers are more active than ever, and it’s not stopping anytime soon. On the contrary, as the digitalization of our lives proceeds, they have more information to collect from public records, marketing databases, scraped web content, or different third-party sources. Then, they can sell it for advertising, profiling, and background checks, among other purposes. Not to mention fraud and identity theft.

The best way to approach this issue is by working with a professional data removal service. And in this space, two names often come up – Incogni and Optery. If you’re comparing, you’re likely asking: Does it really work? Is it legitimate? And which one is better?

The comparison below will walk you through differences without hype, just structure and substance.

Incogni vs Optery at a Glance

Removal Model

The structural difference is pretty clear: Incogni focuses on continuous suppression, while Optery centers on visible exposure management with plan-dependent automation depth.

Broker Coverage

Broker count alone is, of course, essential, but it doesn’t tell the full story; the type of brokers matters as well, if not more.

Incogni reports outreach to more than 420 brokers, including both public people-search websites and private databases involved in marketing, recruitment, background checks, and profiling. Many private brokers operate behind the scenes, don’t provide searchable listings, and are hard to reach for an individual. But addressing these entities targets the backend of the data trading environment directly.

Optery advertises coverage of up to 640+ brokers. However, maximum coverage requires higher-tier plans. Its tools are especially effective with public-facing sites, where listings can be easily identified and tracked. There’s not much Optery does with private brokers.

So, even though Optery may seem to have broader coverage, Incogni’s inclusion of private databases suggests deeper suppression.

Long-Term Performance

A removal service is only as strong as its follow-up system.

The problem of data reappearance is very common in the industry. Databases refresh regularly, and deleting your personal information once doesn’t guarantee that it won’t come back.

Incogni’s requests rely on applicable privacy laws and regulations where possible, and then the service follows up on non-responsive brokers and continuously monitors the web for your data. The whole process was independently assessed by Deloitte, which confirmed it works as promised.

Optery’s effectiveness is evident in public listing removals. In this case, users can easily confirm deletion and monitor progress through the dashboard. Ongoing protection and monitoring are available but depend on the subscription tier. If your concern is mainly with the public visibility of your data, Optery will be fine. 

However, for users seeking long-term, diverse protection, Incogni will bring better effects.

Pricing

Pricing Breakdown: Incogni (2026)

Incogni keeps its pricing structure clear and simple, focusing on automation and comprehensive removal.

Incogni doesn’t fragment its access across tiers – its base plan already includes its entire broker coverage as well as the recurring removal system. No free options are available, but there is a 30-day money-back guarantee for risk-free testing. Family plans expand coverage to more people in a bundle.

For American users, Incogni also offers the Protect plan, which combines its services with NordProtect – it costs $41.48/month or $20.74/month when billed annually.

Additionally, you will find Incogni in the Surfshark One+ bundle, with prices starting at $4.19/month when billed biannually.

Pricing Breakdown: Optery (2026)

Though the number of plans is similar, Optery’s pricing structure is more tiered and modular.

Optery offers a free option that includes an exposure report with dashboard access, links, and screenshots, but no automated removals. The next plan, Core, adds automated removals for a limited set of brokers. The Extended tier expands coverage and adds recurring reports with screenshots. With the top tier, Ultimate, you can reach the highest number of sites – 640+ or 400+ without enabling Expand Reach. You also get unlimited custom removals and quarterly detailed reporting.

Transparency and Reporting

Both services provide their users with dashboards, but the focus differs significantly.

Incogni tracks removal requests, responses, and status updates. As a lot of private brokers don’t publish their listings, screenshot-style confirmation isn’t always possible. Transparency centers on process reporting and tracking.

Optery highlights exposure visibility. You will see the discovered public listings, often with links or screenshots that provide clear confirmation. This approach is especially (and only) useful in the case of public-facing data.

Reputation

Incogni underwent Deloitte’s limited assurance assessment, which confirmed that its processes all work as described. This type of third-party verification is extremely unusual in this industry. Moreover, its Editors’ Choice awards from PCMag and PCWorld, alongside reviews from authorities like TechRadar and Cybernews, make it a strong contender in the privacy protection space.

Optery, on the other hand, doesn’t have any independent verification, but it has been reviewed by industry authorities like PCMag and TechRadar that praise its visibility-focused approach. 

Final Words: Incogni For the Win in 2026

Both Incogni and Optery can be valid data broker removal service choices in 2026. It all depends on what you need.

Optery is strong if you want insight into publicly visible listings and seek visual confirmation of removals.

However, in this 2026 comparison, Incogni ranks as the stronger overall data removal service. This is particularly because of its sustained backend suppression and verified operational processes that were also recognized by industry authorities. 

FAQ

Apple may be preparing to introduce an entirely new high-end laptop later this year.

According to Bloomberg’s Mark Gurman, the company is working on a premium “MacBook Ultra” featuring an OLED display and touchscreen support. This could potentially mark one of the biggest changes to the MacBook lineup in years.

The device is expected to sit above the current MacBook Pro models, rather than replace them. Gurman suggests Apple is positioning the laptop as a new top-tier option. As a result, it is expanding its Mac lineup with a more expensive, flagship-tier machine aimed at users who want the very best hardware.

Anthropic executives allege that current customers and prospective ones have been demanding new terms and even backing out of negotiations since the US Department of Defense labeled the AI startup a supply-chain risk late last month, according to court papers that also revealed new financial details about the company.

Hundreds of millions of dollars in expected revenue this year from work tied to the Pentagon is already at risk for Anthropic, the company’s chief financial officer, Krishna Rao, wrote in a court filing on Monday. But if the government has its way and pressures a broad range of companies from doing business with the AI startup, regardless of any ties to the military, Anthropic could ultimately lose billions of dollars in sales, he stated. Its all-time sales, since commercializing its technology in 2023, exceed $5 billion, according to Rao.

Anthropic’s revenue exploded as its Claude models began outperforming rivals and showing advanced capabilities in areas such as generating software code. But the company spends heavily on computing infrastructure and remains deeply unprofitable. Rao specified that Anthropic has spent over $10 billion to train and deploy its models.

Anthropic chief commercial officer Paul Smith provided several examples of partners who have privately raised concerns to the AI startup in recent days. He said a financial services customer paused negotiations over a $15 million deal because of the supply-chain label, and two leading financial services companies have refused to close deals valued together at $80 million unless they gain the right to unilaterally cancel their contracts for any reason. A grocery store chain canceled a sales meeting, citing the supply-chain-risk designation, Smith added.

“All have taken steps that reflect deep distrust and a growing fear of associating with Anthropic,” Smith wrote.

The executives’ comments are part of statements from six Anthropic leaders in support of a preliminary order that would allow the San Francisco company to continue doing business with the Department of Defense until lawsuits about the supply-chain-risk issue are resolved.

Anthropic has sued the Trump administration in two courts. A lawsuit filed in San Francisco federal court on Monday alleges the government violated the company’s free speech rights. A separate case filed Monday in the federal appeals court in Washington, DC, accuses the Defense Department of unfairly discriminating and retaliating against Anthropic.

The company is seeking a hearing as soon as Friday in San Francisco for a temporary reprieve. The legal battle and sales fallout follows a weeks-long dispute between Anthropic and the Pentagon over the potential use of AI technologies for mass domestic surveillance and autonomous lethal weapons. Anthropic contends AI is not yet capable of safely undertaking the tasks, while the Pentagon wants the right to make that judgment on its own.

By law, the supply-chain designation prevents a narrow set of companies that do business with the Pentagon from incorporating Anthropic into their systems. But Defense secretary Pete Hegseth has cast a wider net. He posted on X late last month that “effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic.”

Rao wrote that the Pentagon reinforced the message by reaching out to several startups about their use of Claude, which he said he learned had happened from speaking with an investor that Anthropic and the smaller companies all share. They “have grown worried and uncertain about their ability to use Claude,” Rao wrote.

The Pentagon declined to comment on the lawsuits and did not immediately respond to a request for comment about Rao’s allegation about the outreach.

More than 30 OpenAI and Google DeepMind employees filed a statement Monday supporting Anthropic’s lawsuit against the U.S. Defense Department after the federal agency labeled the AI firm a supply-chain risk, according to court filings.

“The government’s designation of Anthropic as a supply chain risk was an improper and arbitrary use of power that has serious ramifications for our industry,” reads the brief, whose signatories include Google DeepMind chief scientist Jeff Dean.

Late last week, the Pentagon labeled Anthropic a supply-chain risk — usually reserved for foreign adversaries — after the AI firm refused to allow the Department of Defense (DOD) to use its technology for mass surveillance of Americans or autonomously firing weapons. The DOD had argued that it should be able to use AI for any “lawful” purpose and not be constrained by a private contractor.

The amicus brief in support of Anthropic showed up on the docket a few hours after the Claude maker filed two lawsuits against the DOD and other federal agencies. Wired was first to report the news.

In the court filing, the Google and OpenAI employees make the point that if the Pentagon was “no longer satisfied with the agreed-upon terms of its contract with Anthropic,” the agency could have “simply canceled the contract and purchased the services of another leading AI company.”

The DOD did, in fact, sign a deal with OpenAI within moments of designating Anthropic a supply-chain risk — a move many of the ChatGPT maker’s employees protested.

“If allowed to proceed, this effort to punish one of the leading U.S. AI companies will undoubtedly have consequences for the United States’ industrial and scientific competitiveness in the field of artificial intelligence and beyond,” the brief reads. “And it will chill open deliberation in our field about the risks and benefits of today’s AI systems.”

The filing also affirms that Anthropic’s stated red lines are legitimate concerns warranting strong guardrails. Without public law to govern AI use, it argues, the contractual and technical restrictions developers impose on their systems are a critical safeguard against catastrophic misuse.

Many of the employees who signed the statement also signed open letters over the last couple of weeks urging the DOD to withdraw the label and calling on the leaders of their companies to support Anthropic and refuse unilateral use of their AI systems.

Nvidia is planning to launch an open-source platform for AI agents, people familiar with the company’s plans tell WIRED.

The chipmaker has been pitching the product, referred to as NemoClaw, to enterprise software companies. The platform will allow these companies to dispatch AI agents to perform tasks for their own workforces. Companies will be able to access the platform regardless of whether their products run on Nvidia’s chips, sources say.

The move comes as Nvidia prepares for its annual developer conference in San Jose next week. Ahead of the conference, Nvidia has reached out to companies including Salesforce, Cisco, Google, Adobe, and CrowdStrike to forge partnerships for the agent platform. It’s unclear whether these conversations have resulted in official partnerships. Since the platform is open source, it’s likely that partners would get free, early access in exchange for contributing to the project, sources say. Nvidia plans to offer security and privacy tools as part of this new open-source agent platform.

Nvidia did not respond to a request for comment. Representatives from Cisco, Google, Adobe, and CrowdStrike also did not respond to requests for comment. Salesforce did not provide a statement prior to publication.

Nvidia’s interest in agents comes as people are embracing “claws,” or open-source AI tools that run locally on a user’s machine and perform sequential tasks. Claws are often described as self-learning, in that they’re supposed to automatically improve over time. Earlier this year, an AI agent known as OpenClaw—which was first called Clawdbot, then Moltbot—captivated Silicon Valley due to its ability to run autonomously on personal computers and complete work tasks for users. OpenAI ended up acquiring the project and hiring the creator behind it.

OpenAI and Anthropic have made significant improvements in model reliability in recent years, but their chatbots still require hand-holding. Purpose-built AI agents or claws, on the other hand, are designed to execute multiple steps without as much human supervision.

The usage of claws within enterprise environments is controversial. WIRED previously reported that some tech companies, including Meta, have asked employees to refrain from using OpenClaw on their work computers, due to the unpredictability of the agents and potential security risks. Last month a Meta employee who oversees safety and alignment for the company’s AI lab publicly shared a story about an AI agent going rogue on her machine and mass deleting her emails.

For Nvidia, NemoClaw appears to be part of an effort to court enterprise software companies by offering additional layers of security for AI agents. It’s also another step in the company’s embrace of open-source AI models, part of a broader strategy to maintain its dominance in AI infrastructure at a time when leading AI labs are building their own custom chips. Nvidia’s software strategy until now has been heavily reliant on its CUDA platform, a famously proprietary system that locks developers into building software for Nvidia’s GPUs and has created a crucial “moat” for the company.

Last month The Wall Street Journal reported that Nvidia also plans to reveal a new chip system for inference computing at its developer conference. The system will incorporate a chip designed by the startup Groq, which Nvidia entered into a multibillion-dollar licensing agreement with late last year.

Paresh Dave and Maxwell Zeff contributed to this report.