gshc2020.com

It’s still not clear when Project Helix will actually make it to market, but over this last weekend, there were rumblings that Microsoft was thinking about pulling back from Steam. Those rumblings came from Jez Corden’s Xbox Two podcast, and Corden wrote out a long post on Windows Central yesterday basically calling it a misunderstanding.

Xbox isn’t likely to pull its games from Steam any time soon, but with the cost of hardware being what it is, I could see the company walking back on the open nature of Project Helix. Launching an open PC platform is a risky move when hardware costs keep going up. Just look at the Steam Machine – due to its nature as an open platform, Valve couldn’t reasonably subsidize the hardware, and it led to that mini PC getting a four-digit price tag.

Project Helix, as it was pitched back in March, would basically be a Windows-flavored version of what the Steam Machine is doing. But because it’s going to have ‘next-gen’ pressures, the hardware will probably be significantly more powerful than what you’ll find in the GabeCube. If Microsoft isn’t able to subsidize the cost – and if it's an open platform, it won’t be able to – it’s likely that you’re going to have to pay gaming PC money to get the next Xbox.

I’m not sure how many people would be willing to spend one or two thousand bucks on a next-generation gaming console, especially if Sony isn’t also charging that much. It feels like a foregone conclusion right now that the $500 console is dead, but even if it is, I’m not sure that the PS6 would actually reach $1,000. Just last month, industry analyst Piers Harding-Rolls said on Twitter that he didn’t think the Steam Machine’s high price necessarily means the next PlayStation would hit a thousand bucks, mostly due to Sony’s more robust supply chain.

Either way, it’s a little early for any of this to be anything more than just speculation. Right now, it looks like the PS6 isn’t going to be coming out until 2028 anyways, and there’s a possibility that the RAM crisis will have died down a little before then. But if hardware is still in short supply when Project Helix and the PS6 come out, I almost want Microsoft to double down on making its next-gen console a gaming PC.

If Microsoft really wants to court PC gamers, it makes a lot of sense to have Project Helix as a sort of first-party halo (no pun intended) product. Ideally, Microsoft would just make the console/PC hybrid as close to a top-end rig as possible and position it as the premium option. Then, Microsoft could just leave the entry-level to third-party manufacturers or PC builders. After all, part of the magic of PC gaming is its scalability, so you wouldn’t need to fork over whatever Microsoft is asking for Helix, if you’re willing to build something yourself.

It wouldn’t even be the first time that Microsoft took this kind of approach. Outside of Xbox, this is exactly the mindset behind the Surface lineup of PCs. Microsoft makes a beautiful laptop as a model of what the company thinks PCs should be, but still allows Windows and all of its software to run on other hardware. I’ve even joked to friends in the past that it was only a matter of time before Microsoft made the Xbox Surface, and that’s kind of what Project Helix looks like, at least right now. Not a tablet, to be clear, but a sleek, high-end device.

Even though there are rumors that Microsoft is going to keep supporting discs, it still feels like the death of physical games is looming around the corner, so Project Helix going with an open PC platform would at least allow you to buy games on GoG. And, there, you’ll have a more convincing illusion of owning your games than just buying them on the Xbox store. Plus, if you’re anything like me, where you own hundreds of games across various PC storefronts, you’ll have an instant library on day one.

The problem with Helix being an open platform, though, is the UX. One of the most frustrating things about the Asus ROG Xbox Ally X continues to be the Xbox Full Screen Experience. Microsoft has done a lot to improve it since I reviewed the handheld last year, but to this day it’s still sometimes buggy and unresponsive, especially on the base Xbox Ally. I frequently have problems where I go to close a game, and it just doesn’t work, leading me to mashing the X button in Task View until the game eventually closes.

I do think the Xbox Full Screen experience works pretty well on a TV, and would make a lot of sense on a console, but only if Microsoft puts in the work to actually polish it. Luckily, the company has plenty of time to tighten the screws before the next-gen consoles launch, and Project Helix will likely be powerful enough to brute force some of the issues the handhelds have with the OS. Ideally, the FSE should be as intuitive as SteamOS by the time the console hits living rooms.

However, with the way hardware pricing is going, and with how CEO Asha Sharma seems intent on bringing back Xbox’s console identity, it’s possible that Microsoft pulls a 180 on Project Helix, making it a traditional follow-up to the Series X. It would make sense to funnel people into the Xbox store where Microsoft could make a bigger cut of each game sold. Maybe that’s even the path Xbox should take. But I really think Project Helix would be losing something special by just being a sequel to the Xbox Series X, rather than taking a stand as Microsoft’s example of what a gaming PC should be.

Jackie Thomas is the Hardware and Buying Guides Editor at IGN and the PC components queen. You can follow her @Jackiecobra %!s()

Fake shifts were slowing it down

Mercedes-Benz

It may have beaten the competing Hyundai around the Nürburgring, but this new AMG gets a feature notably inspired by that company's N division: internal combustion simulation. In its AMGFORCE Sport+ drive mode, sounds recorded from the outgoing CLA45's turbo four can be heard, complete with pops and bangs. Those live-generated sounds are paired with simulated gearshifts, a trope started with the Hyundai Ioniq 5 N and since implemented in the 6 N as well as the refreshed Porsche Taycan. As you can see from the video below, however, the car was in Race mode for the hot lap, which turns off those shifts and noises in favor of AMG's more spaceship-y sound and traditional EV acceleration.

The CLA45 goes from zero to 62 mph in 2.7 seconds, tops out at 168 with the AMG Dynamic Plus package and can indeed be seen approaching that very speed on the Nurburgring's Döttinger Höhe back straight. It's also got active aerodynamics like a pop-up rear spoiler that are pretty cool to see on such a small car. But what really made the difference on the lap was the car's Predictive Performance Manager system:

The system analyzes the track layout and optimizes energy deployment based on the circuit's topography and the current driving situation. As a result, the available power is delivered precisely where it has the greatest impact on lap time. On demanding circuits such as the Nürburgring Nordschleife, performance is not achieved through continuous maximum power output, but through the intelligent distribution of available energy throughout the entire lap. This is exactly where the PPM comes into play, creating the foundation for both fast and highly repeatable lap times. The driver can choose between two operating strategies: "SPRINT" for achieving the fastest possible lap time, and "RACE" for multiple laps with maximum consistency and sustained performance.

As cool as the new CLA45 is, Mercedes has unfortunately confirmed that it won't be sold in the United States. We may get the powertrain in the new-gen GLA, at least, and AMG versions of the electric C-Class and GLC will use even more-powerful tri-motor setups. %!s()

The company reported revenues of $6.9 billion, up 1 percent from its predecessor company a year ago, while net earnings fell to $41 million, or four cents per share after $57 million, or eight cents a share, a year ago.

Paramount’s TV media segment, which includes CBS and its cable networks, continued to fall, dropping 9 percent year-over-year to $3.1 billion, while direct-to-consumer revenue rose 9 percent to $2.5 billion. Studio revenue rose 16 percent to $1.3 billion, thanks to content licensing deals and TV production arm as well as the success of Scary Movie.

Revenue at Paramount+ rose 16 percent to $2.1 billion, as the company said Q2 was its best quarter for retention ever on Paramount+ thanks to Dutton Ranch, UFC and the FIFA World Cup. The streaming services added 2 million new subscribers to hit 81.6 million worldwide, up 6 percent year over year. 

Paramount is raising its full-year outlook range to $3.8 billion to $3.9 billion in adjusted EBITDA and now expects free cash flow conversion of at least 10 percent. The company added that it had seen “double-digit growth in commitments” across the company after its Upfronts presentation, but did release further specifics.

“We continue to make progress on our transformation at Paramount, and we now expect to deliver over $2.7 billion of run-rate efficiencies by the end of 2026 versus $2.5 billion previously, and continue to expect $3 billion-plus in efficiencies from the Skydance-Paramount combination,” the letter continues.

Paramount won the bidding war for Warner Bros. Discovery in February with its $111 billion offer, after Netflix declined to raise its offer. But the merger is currently on pause, as Paramount has since been hit with several lawsuits, including from 12 state attorneys general who claim the acquisition will kill competition in the film and television industries. The company has agreed to delay the billion takeover of Warner Bros. Discovery until June 1, 2027, or until the antitrust lawsuits brought by state attorneys general are resolved. Ellison has told staff he remains “highly confident” in the completion of the merger.

But just after earnings were released Tuesday, the judge overseeing the states’ lawsuit scheduled the trial to start in March, rather than November as the studio had requested. Starting Oct. 1, Paramount will have to pay a ticking fee of roughly $7 million per day to Warners shareholders until the deal closes.

“As it relates to the ongoing litigation, we’re absolutely open to finding a solution out of court, but we also really believe that we’ll win at trial. We believe that the facts and the law are on our side, and the trial date was just set for for March of next year. And as it relates to the financing, all that is in place, there’s nothing at risk, and so we’re confident we’ll close the transaction, and we’re working towards that as fast as we possibly can,” Ellison said on the earnings call.

As for how much this will cost, CFO Dennis K. Cinelli said the company incurs costs beyond Sept. 30, with monthly bridge fees of $8 million to $9 million a month, in addition to an additional bridge commitment fee due in June 2027. In total, this adds up to about $190 million dollars if the merger doesn’t close until June. This is in addition to the ticking fee, which is 25 cents per share per quarter or about $650 million dollars per quarter, and would be funded at close through additional equity.

“We ended the quarter at $1.6 billion in cash, $3.2 billion of undrawn revolver capacity. This is sufficient to fund the business, our dividend, transaction-related costs through the extended timeline,” Cinelli said.

The earnings also come the same day that an op-ed from Ellison appeared in The New York Times, in which the executive argued that his political leanings and proposed takeover of CNN are the root cause of the lawsuits filed by a dozen state attorneys general and the Writers Guild of America.

“I believe this fight is not really about market share … I believe a plainer worry sits beneath the briefs and the news releases: the news. The issue is whether I can be trusted as a steward of Warner’s CNN. There has been speculation about my politics, my loyalties, my intentions,” Ellison wrote. “Unfortunately, I can’t give anyone a view into my heart and mind, but I can share this: I have regularly voted for candidates of both parties; I hold some views that would be called conservative and others that would be called liberal, just like most Americans; and when it comes to our news operations, I do not aspire to lead these companies to bend their newsrooms to my views. I believe that news should be based on facts and truth.” %!s()

The deal will see Hearst take full control of the media company formerly known as A+E Networks, which owns brands like A&E, History and Lifetime, including cable TV channels, content studios, and streaming offerings. A+E has also invested in companies like Vice, Propagate Content, Atlas Obscura and Philo TV.

A+E will become a part of Hearst’s entertainment group upon closing later this year.

“We thank our Disney colleagues for decades of successful partnership,” said Steven R. Swartz, president and CEO of Hearst. “We look forward to supporting Paul Buccieri and A+E Global Media’s leadership team as they continue to make must-see programs and innovate around the great HISTORY, Lifetime and A&E brands.”

Paul Buccieri has led A+E since 2018.

“In a media environment defined by fragmentation, A+E Global Media’s advantage is the strength and versatility of our brands, our strong partnerships and our vast library of owned assets,” added Buccieri, president and chairman of A+E Global Media. “As we continue extending our storytelling globally across all platforms with IP that travels to every screen and form-factor, we believe we are well suited for whatever opportunities may come next. I also want to express my deepest thanks to Hearst, The Walt Disney Company and to our board members — both recent and past — for their guidance and support over the years.”

The company was founded in 1984 (the Arts & Entertainment Network was its original cable brand), with Hearst and ABC each owning a slice alongside other investors. NBC obtained a minority stake in 1993. The company expanded with an acquisition of Lifetime Networks in 2009. Lifetime was also a JV of Disney and Hearst.

Disney and Hearst had been equal partners in A+E since 2012, when they bought out NBCUniversal’s minority stake. The company had begun to explore its options a year ago, hiring Wells Fargo to market it. Since then, of course, Comcast spun out its cable channels into Versant while Warner Bros. Discovery has inked a deal to sell itself to Paramount Skydance. %!s()

According to SaferAI’s evaluation, which the nonprofit ran via Z.ai’s public API, GLM-5.2 refused none of the offensive cyber or dual-use biology tasks it was given. By comparison, Claude Opus 4.7 “refused so consistently that SaferAI could not complete CyberGym on it at all.” (CyberGym is a benchmark that evaluates cybersecurity capabilities. OpenAI used it in the evaluation that preceded last month’s Hugging Face breach.)

It’s a stark reminder of what some critics have warned for years: that open-weight AI models could put highly capable AI into the hands of potential attackers, with no way to police how they use the technology once they download the weights. With open-weight models rapidly approaching the capabilities of the world’s leading AI systems, the debate is moving from whether they can compete to how society manages risks once they are released. 

“The frontier of capability is not the frontier of risk, and so we do have to take into account the state of the mitigations as well to assess the risk properly,” Henry Papadatos, executive director of SaferAI, told TechCrunch.

While Z.ai could apply safety measures to its hosted API, those protections become unenforceable once someone runs the weights on their own hardware, where they can remove or modify any safeguards, fine-tune the models, or change system prompts. 

Frontier developers like OpenAI and Anthropic tend to rely on safeguards like classifiers, refusal training, and API-level controls to limit dangerous cyber and biological assistance. 

Those measures are far from foolproof: jailbreaks routinely bypass protections on deployed models. Far.ai, an AI safety nonprofit, found hundreds of universal jailbreaks — defined as reusable keys that succeed on most harmful requests — in frontier models like xAI’s Grok 4.5 and Google DeepMind’s Gemini 3.1 Pro. According to the report, jailbreaks succeed when attackers combine multiple manipulation techniques — including roleplaying, authority impersonation, fake conversation history, and follow-up prompts — to amplify weak points in a model’s defenses. 

But the safeguards in place for closed models don’t work at all on open-weight models, which are designed to run on any infrastructure with any set of safeguards — or lack thereof.

“The objective should clearly be that the good capabilities — the safe ones — are accessible to anyone, and then we try to remove the bad ones, even in an open source fashion,” Papadatos said.

One technique Papadatos noted could help is called “pre-training data filtering,” which is when an AI company removes offensive cybersecurity information from their training data and then trains the model on the curated dataset. 

Some research suggests this can reduce hazardous biological knowledge without harming overall model performance. However, for cybersecurity, data filtering is much less practical. 

It’s difficult to train a general model that excels at coding but isn’t also a good hacker. Because coding has become AI’s biggest moneymaker, developers face pressure to keep improving those capabilities even as they search for ways to limit misuse. 

Because of that, frontier developers have increasingly relied on other mitigations instead. One approach has been to selectively restrict the kinds of cybersecurity assistance models will provide. Anthropic’s Opus 5, for example, can search for vulnerabilities in uncompiled source code, but not compiled software, per the model’s system card. The reasoning is that this makes it harder to use Opus 5 for offensive purposes. 

Others include rigorous pre-deployment safety evaluations, publishing risk assessments, and withholding model weights if a system is perceived as too dangerous. 

In GLM-5.2’s case, SaferAI says Z.ai didn’t publish a safety framework, pre-deployment testing commitments, or risk assessment for the model. TechCrunch has asked Z.ai whether it conducted internal or third-party frontier safety evaluations before release, but did not receive a response. 

Chinese leaders have increasingly acknowledged the risks of advanced AI. At the World AI Conference last month, Chinese President Xi Jinping emphasized the importance of open-weight models, while also stressing the necessity of ensuring AI remains a tool under strict human control. 

Graham Webster, who studies Chinese AI policy at the Stanford Cyber Policy Center, told TechCrunch that China has robust regulations governing AI, but those rules have historically focused on politically sensitive content, misinformation, and social stability rather than catastrophic AI risks like offensive cyber capabilities and biological misuse. 

“U.S. AI thinkers are, in general, more concerned with this existential catastrophic [idea] than the Chinese community,” Webster said, adding that many Chinese policy researchers believe that if there’s truly going to be a novel frontier risk, American companies will likely encounter it first. 

“The Chinese system has confidence that they control the use of these technologies inside China,” Webster continued. “Being online in China is something you do attributed to your real name, and companies can be held accountable, users can be held accountable.”

Webster mused that the same mechanism that model providers use for refusing to engage on certain political topics can potentially be tweaked to make sure models refuse to complete offensive cyber attacks or won’t deliver adverse biological engineering outcomes. He added that because Chinese companies tend to coordinate with regulators behind the scenes, it can be tough to know what internal testing they’re conducting before release.

Advocates of open-weight AI argue that releasing the weights is important for cybersecurity because it allows companies defend themselves against attacks — Hugging Face relied on GLM-5.2 to defend itself against OpenAI’s breach — and because it allows them to better prepare for future threats if they know what’s coming.

“The same systems that helped stop an AI-powered cyberattack can now help defend against millions of cyberattacks every day, while helping us identify and fix vulnerabilities before attackers exploit them,” Clem Delangue, CEO of Hugging Face, said this week in a social media post.

Papadatos said that benefit is often overstated, and doesn’t mean “we should open-source dangerous capabilities.”

“The main point in my mind is that we shouldn’t just accept that dangerous capabilities are easily accessible by anyone anywhere,” he said, stressing that he believes the industry should be striving for only making the “good capabilities” easily accessible. By default attackers adopt new tools faster than defenders do. For example, a ransomware group can change its methods in a week. A hospital cannot.”

When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence. %!s()