Chris Fall, the director of the Center for AI Standards and Innovation (CAISI), has resigned, the agency confirmed to multiple news outlets.
He was appointed just three months ago after the last appointee, Collin Burns, left in less than a week, The Washington Post reported at the time. Burns was reportedly “pushed out” of the job in April because he previously worked for Anthropic and the Trump administration had been battling with the company, sources told the Post.
No reason was given for Fall’s departure. Prior to leading CAISI, Fall was the director of the Department of Energy’s Office of Science during the first Trump administration and had been the acting director of the DOE’s Advanced Research Projects Agency-Energy. He worked in the DOE’s Office of Naval Research (ONR) prior to that.
Before Burns and Fall, the agency was led by venture capitalist David Sacks, whose title at the time was White House AI and crypto czar. Sacks stepped down in March.
Advertisement
CAISI, which operates under the National Institute of Standards and Technology, is the primary organization for developing technical standards and testing methods for AI models as well as assessing cybersecurity risks. Yet it was not the agency at the center of the most recent model-risk brouhaha.
That occurred in June when the U.S. Commerce Department invoked an obscure export control directive that effectively forced Anthropic to pull its Mythos and Fable models from the market. The ban was lifted by the end of the month, when Secretary of Commerce Howard Lutnick said he was satisfied with Anthropic’s safety plans.
Earlier this month, the White House also signed an executive order for a new AI safety oversight program called “Gold Eagle” that creates a clearinghouse for cybersecurity vulnerability coordination. A host of federal organizations were named as part of the program, including the Commerce Department and Department of Homeland Security. But, as CNBC pointed out, CAISI was not among the federal organizations mentioned.
Meanwhile, after Anthropic’s models were freed from the ban, Google DeepMind CEO Demis Hassabis began calling for the creation of an independent, industry-run standards body to regulate frontier AI modeled after FINRA — the same sort of mission that CAISI was formed to tackle.
Advertisement
Fall’s resignation also follows this weekend’s handwringing over Chinese AI lab Moonshot’s new version of its open model Kimi, which performed competitively against flagship frontier models. The administration was weighing efforts to somehow ban Chinese open models, Axios reported. This sparked immediate debate and outrage over the weekend, including from Sacks, who argued that regulations shouldn’t be used as a protectionism strategy for U.S. proprietary AI labs.
While CAISI has released a few reports on the capabilities of Chinese open-weight models Z.ai’s GLM-5.2 and DeepSeek V4 Pro, it hasn’t talked much about its processes for testing. (Open weight means these models can be publicly downloaded and run locally, but its training code and datasets are not available). Since July 9, TechCrunch has sent multiple inquiries to both the DoC and NIST about how its LLM evaluations work and has not received a response.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.
AI models will do just about anything to complete the task you ask, including cheating to get there, according to new cybersecurity evaluations from the UK government’s AI Security Institute (AISI). The group found that leading models often take shortcuts to achieve a particular result and then misrepresent how they obtained that result. And they won’t always admit it when asked. “Every model we have tested for this behaviour attempted to cheat,” AISI said in a blog post on Tuesday. “Models did not reliably report this behaviour when asked, and often did not reason about it in their chain-of-thought, suggesting that detecting cheating will likely require robust monitoring methods.” Infractions included searching the internet for the answer, bypassing sandbox network restrictions, probing the evaluation harness, attacking a system other than the target, and guessing an answer. Cheating in this manner – employing a workaround or gaming a reward function to score better on a benchmark test, for example – has been widely documented by machine learning researchers. It doesn’t necessarily imply malicious intent, AISI said, but it’s nonetheless troublesome because it can produce misleading assessments of model capabilities. When AISI conducted evaluated five leading models, it found that all of them cheated. The results were as follows: GPT-5.4 cheated 67 times in 475 test runs (14.1 percent). GPT-5.5 cheated 54 times in 475 test runs (11.4 percent). GPT-5.6-Sol cheated 60 times in 475 test runs (12.6 percent). Claude 4.7 Opus cheated 43 times in 475 test runs (9.1 percent). Claude Mythos Preview cheated 37 times in 475 test runs (7.8 percent). Asking models whether they cheated or did anything wrong proved an unreliable auditing mechanism because the models didn’t always admit wrongdoing. “In our experiments, models did not consistently acknowledge attempted cheating when asked, and described it as wrong less than 50 percent of the time,” said AISI. Existing vetting methods, such as self-reporting and chain-of-thought logs, proved similarly dicey because models don’t always report their chain-of-thought. And there were instances where a model would consider whether a proposed action amounted to cheating and then decided to take the action anyway. Given the absence of reliable model cheating detection methods, AISI warns that its current approach – manual review coupled with LLM monitoring – may not be sufficient to catch deception, particularly as models become more sophisticated. “A more fundamental fix would be to train the models not to cheat in the first place – but given this kind of behaviour was reported in frontier models more than a year ago, robustly aligning it away may not be easy,” AISI concludes. ®
Recording artist, DJ and entrepreneur Diplo invested in Copper. Photo via BusinessWire
Seattle’s Copper has landed a high-profile new backer as it looks to accelerate growth of its consumer rewards platform, announcing Tuesday that Grammy-winning artist, DJ and entrepreneur Diplo has invested in the company.
Financial terms of the investment were not disclosed.
“I’m always looking for things that actually make sense for people,” Diplo said in a statement. “Copper’s one of those — you’re already on your phone, you’re already spending money, and this gives something back. That’s real.”
Copper says more than 4 million members use its platform to earn money through mobile games, cash-back offers and purchases.
Copper CEO Eddie Behringer, who previously co-founded Snap! Raise, said the company is building an alternative to consumer apps that monetize users’ attention.
Advertisement
“Most consumer apps are designed to take more from the user — more time, more money, more attention,” Behringer said in a LinkedIn post. “At Copper, we’re building the opposite.”
Founded in 2019, Copper originally launched as a banking app for teenagers. GeekWire covered the startup in 2022 after it raised $29 million in funding to expand into investing products, at a time when the company had nearly 1 million users.
The startup has since evolved into a broader consumer rewards platform. Copper has raised $42 million to date and recently ranked No. 2 among the Pacific Northwest’s fastest-growing companies in Deloitte’s Technology Fast 500 rankings, based on three-year revenue growth.
Diplo, whose real name is Thomas Wesley Pentz, has built a business portfolio that extends beyond music, investing in technology and consumer startups while launching ventures such as Diplo’s Run Club, a series of 5K races paired with music festivals.
Advertisement
He’s a three-time Grammy winner, and has collaborated with artists like Labrinth and Sia as part of the musical group LSD and worked with musician Mark Ronson on Silk City. He’s also the founder of record label Mad Decent.
In 2024, Copper discontinued its banking services following the collapse of fintech infrastructure provider Synapse, forcing the startup to pivot away from its original business. “Despite our prior planning, this event has forced us to close banking accounts much sooner than anticipated,” Behringer wrote at the time.
The company has since rebuilt around its rewards platform, which it says now serves millions of users.
Behringer said that the company’s mission was always about helping families improve their financial lives.
Advertisement
“As household costs rose, we saw an even bigger opportunity to help the person making everyday spending decisions earn more from the things they were already doing—from buying groceries to shopping in-store and spending time on their phone,” Behringer tells GeekWire via email. “Diplo’s investment is meaningful validation of how far that evolution has come.”
The company has partnered with Pangram for the new tools.
Substack
Substack has launched a new AI detection tool in partnership with Pangram. This will allow readers to scan Substack content for an assessment of how much of the material was written by AI. The tool can be used on text longer than 100 words that was published beginning today. Substack is also adding a new statement space for creators to explicitly share if and how they used AI for their content. The AI detection capabilities are available starting today on web and iOS, with Android support to come.
The blog post announcing the feature is surprisingly spicy. There’s a dig at LinkedIn about the presence of AI-generated content on that service and it dubs attempting to create feigned human connection with AI slop “Claudefishing.” Throwing shade is a risky maneuver here, because even the best tools for identifying gen-AI can’t guarantee a correct assessment. The Atlantic dug deeper into just how accurate AI detection tools, Pangram in particular, can be. Spoiler: they’re far from perfect.
Substack did acknowledge in the post that there are limits to what Pangram can detect and hinted at some other features it is considering around AI content and preferences. It emphasized that these new measures are aimed at setting expectations for readers, summing up its stance as “people should know what they’re getting.”
Google released three new AI models on Tuesday, all built on Gemini 3.5 Flash. The new models are more token-efficient, faster and more reliable across the board. The tech giant also provided an update on the much-anticipated Gemini 3.5 Pro and what’s to come after.
A new AI model from the likes of Google, Anthropic or OpenAI is released seemingly every week, with the latest, ChatGPT-5.6, released earlier this month. Google’s latest releases aren’t flagship models compared to what’s on the horizon, but each has its place, including a new model solely focused on cybersecurity.
Here’s what’s new in the latest Gemini models from today’s announcement.
Gemini 3.6 Flash
Google called 3.6 Flash its “workhorse” model that’s now better at coding, knowledge work and multimodal performance. It also promises reduced token usage by up to 17%, and at a lower cost per token versus its predecessor, 3.5.
Google says it built the model based on both developer and customer feedback. A series of benchmarks shows 3.6 Flash’s gains in performance and average tokens per task compared to its predecessor.
Advertisement
Gemini 3.5 Flash-Lite
Google’s fastest and most cost-effective model can deliver 350 output tokens per second and “significantly” outperforms previous generations when it comes to agentic workflows, according to the blog post.
Like 3.6 Flash, this model now supports computer use as a built-in tool to take on more agentic tasks.
Gemini 3.5 Flash Cyber
3.5 Flash Cyber is a specialty model that prioritizes cybersecurity workflows in order to find and fix vulnerabilities. It works alongside an infrastructure agent called CodeMender to help cybersecurity teams quickly identify and patch issues.
According to a separate article from Google DeepMind, the new model is already finding and fixing bugs in Google’s internal codebases in Android, Chrome and YouTube. This model will initially be limited to governments and trusted partners, but access will expand in the future.
Advertisement
Gemini 3.5 Pro is still on the way
While the three latest models are the primary focus for Tuesday’s announcements, Google gave a brief update to its upcoming flagship AI model, Gemini 3.5 Pro. The model is said to currently be in testing with partners, and it plans to make it available as soon as it’s ready. How long that will take is anyone’s guess, but Google’s also already looking ahead to the next generation of AI, too.
Google says it has already begun pretraining for Gemini 4, which will be released at an undetermined date.
Both Gemini 3.6 Flash and 3.5 Flash-Lite are available starting today for developers in the Gemini API via Google AI Studio and Android Studio and the Gemini app. 3.6 Flash is also available in Google Antigravity, and 3.5 Flash-Lite is rolling out to Google Search.
Blake has over a decade of experience writing for the web, with a focus on mobile phones, where he covered the smartphone boom of the 2010s and the broader tech scene. When he’s not in front of a keyboard, you’ll most likely find him playing video games or watching horror movies.
See full bio
Disclaimer: Unless otherwise stated, any opinions expressed below belong solely to the author.
Given the number of various support schemes provided by the government in Singapore, one might question whether there aren’t too many of them and if it wouldn’t be simpler to simply disburse one cash payment to every eligible person.
After all, it’s not like the government isn’t doing that already, depositing funds directly for GST Vouchers, Assurance Package payments and cost-of-living support, regularly appearing in recent years.
So why bother with CDC Vouchers, which require an entire digital infrastructure to allow their issuance and redemption? Wouldn’t a simple bank transfer be better?
Advertisement
Cash is not king
At least not for everybody, and certainly not for the government, which is using its power to direct the money to specific parts of the economy.
Cash is liquid, and you can do whatever you want with it, including going for a nice day trip to JB to spend it there instead of Singapore. You may also use it for online shopping on one of the many ecommerce platforms, with most of the funds being sent to sellers in China or other countries.
This sees Singapore dollars exiting the domestic economy, benefiting others instead.
More prudent Singaporeans could opt to save it instead, which isn’t terrible in itself, but does keep the funds out of circulation.
Advertisement
Finally, the more reckless consumers could simply waste it on more “sinful” pleasures, still ending up short of money for daily necessities.
The voucher format allows the government to set strict rules on their use: with 50% allocated to shopping in supermarkets and another half to hawkers and smaller, heartland merchants.
This ensures that this pool of money is spent in the most beneficial way and provides the authorities with data on how the money is spent and where.
There is, however, one other purpose they have served very well since their launch six years ago, which is not spoken of.
Advertisement
Going digital—and staying there
CDC Vouchers originated as a COVID-19-era relief scheme, originally issued on paper and directed to the poorest households.
With the pandemic dragging on throughout 2021, the scheme was ultimately expanded to all citizen households by the end of the year and went digital with the launch of the RedeemSG app. Merchants could use it to accept the vouchers by scanning digital QR codes on customers’ phones, instead of dealing with paper.
In parallel, the government launched the Hawkers Go Digital scheme in Jun 2020, with generous subsidies and transaction fee waivers, which were meant to help hawkers adopt digital payments and reduce the risk of spreading the virus.
It was also a good opportunity to prod them to adopt mobile payments, which have become a staple in many countries around the world (most notably China, through its giant superapp WeChat).
Advertisement
The final waiver of the 0.5% fees ended just recently on Jun 30, 2026, after several past extensions. This means that hawkers are now going to have to bear the cost themselves, which might mean that some of them may prefer to return to cash-only payments.
But this is where the CDC scheme comes in.
Throughout the push for digital payments, the critics lamented that many elderly sellers might be struggling with the transition, not being very tech-savvy. What’s more, a skill once developed needs to be kept in use before it falls out of favour. Old habits die hard, after all.
Well, while we might see some return to cash, there is no returning to paper for CDC vouchers.
Advertisement
And because of the scale of the program, currently exceeding S$1 billion annually, hawkers have a strong incentive to keep their RedeemSG app to accept voucher QR codes.
In other words: there’s no exit from the QR era.
Of course, using the CDC app doesn’t force merchants to accept all digital payments, but since they still have to deal with QR codes to accept the vouchers, it provides very useful stickiness, which is going to keep most of them on the digital train.
Read other articles we’ve written on Singaporean businesses here.
Nvidia is hyping up its new Vera Rubin chip system this week, revealing new performance benchmarks for the GPU and CPU combo ahead of rival AMD’s annual product event in San Francisco on Thursday.
During a lengthy technical workshop last week at the company’s headquarters in Santa Clara, California, Nvidia executives boasted to a small group of journalists about the chip system’s increased power and efficiency capabilities. The biggest takeaway: Nvidia, which has long specialized in making GPUs, is increasingly trying to position itself as a supplier of CPUs that can power AI agents.
While GPUs are still the main hardware that companies use to train and run their AI models, the industry’s shift toward more complex, agentic systems has increased demand for CPUs, which can orchestrate data flows, networking, and other software tasks. That’s likely one reason Nvidia has been eager to promote itself as a supplier of complete AI systems rather than just AI chips.
Vera Rubin is Nvidia’s successor to its hybrid superchip system Grace Blackwell and represents the linchpin of its near-term future powering the AI industry. It’s designed to offer one CPU for every two GPUs. In a single Vera Rubin NVL 72 super chip system, there are 36 Vera CPUs for every 72 Rubin GPUs. Nvidia is also selling the Vera CPU as a stand-alone product, and it has reportedly told Chinese customers these could be ready as soon as August.
Advertisement
The CPU chip Nvidia is using for its new Vera Rubin hardware system.
Courtesy of Nvidia
Nvidia executives emphasized that its new Vera Rubin NVL72 racks—a stack of chips packed into a single liquid-cooled platform—are much more “plug-and-play” than some of its earlier products. During a brief tour of a Nvidia data center lab in Silicon Valley, Nvidia executives shared that OpenAI already has one Vera Rubin rack in use.
Nvidia CEO Jensen Huang didn’t make an appearance at the workshop in Santa Clara last week; he was in Japan announcing the chipmaker’s new partnerships with a number of Japanese firms to develop AI for robotics. The briefings were instead led by Ian Buck, Nvidia’s longtime vice president of accelerated computing and the architect behind the company’s CUDA software.
“We’re on a road map to crank out new architectures, not just GPUs but CPUs,” Buck told reporters. “We’re going to keep innovating, because it’s do this or die in Silicon Valley.”
Advertisement
The meetings were held in Huang’s executive briefing center, where multiple desks nearby were piled with bags of Taiwanese snacks that the CEO brought back from his recent trip to Computex, a massive annual semiconductor trade show in Taipei, an Nvidia spokesperson told WIRED.
If you’re a fan of the Ford F-Series pickup trucks, then you’re likely familiar with their technology and design. For example, you may know about the automaker’s legendary Twin I-Beam front suspension. But you might not know that Ford chose to move away from this design on the F-150, not because it was flawed, but because trucks evolved to more modern solutions.
The Twin I-Beam initially gave Ford a way to improve ride comfort while also maintaining the rugged performance that truck buyers were accustomed to. But as pickups began to shift from being used primarily as utility vehicles to everyday drivers, customer expectations began to change. Ford redesigned the F-150, focusing more on precise handling, improved steering, and better control over the truck’s front end. Newer designs like dual A-arm suspension addressed those concerns, leading Ford to move the redesigned 1997 F-150 to a different front suspension design.
Even as the F-150 transitioned from the Twin I-Beam, the design remained one of Ford’s most well-known truck innovations and continues on the F-250 and F-350 trucks to this day. In fact, the Twin I-Beam setup was exclusive to Ford when it was introduced in 1965 and became closely associated with their most popular pickups. Vehicle axles are more complicated than you might think and as other companies used different suspensions for their trucks, Ford’s approach was more forward-thinking. It gave drivers a vehicle with independent front-wheel movement, while also maintaining durable performance.
Advertisement
The evolution of Ford truck suspension
Jetcityimage/Getty Images
As Ford was developing its innovative Twin I-Beam concept, the automaker introduced the Twin Traction Beam in 1980. This suspension was designed for four-wheel-drive trucks like the F-150 and Bronco, using the independent movement of Twin I-Beam while adding components for a driven front axle. Twin Traction helped reduce weight, improve ride quality, and lower the truck’s overall height. This design helped Ford modernize its 4WD lineup at the time of its production.
In the years since the development of the Twin I-Beam, Ford’s approach to truck suspension has continued to evolve. Instead of using a single design for every truck in the automaker’s lineup, the F-150 now features different suspension setups depending on its intended purpose. For example, the standard F-150 has a Hotchkiss-style suspension, with a solid rear axle and leaf springs for both towing and payload. In contrast, the F-150 Raptor features a five-link rear suspension with coil springs, and the all-electric F-150 Lightning uses a fully independent rear suspension.
Advertisement
This design evolution extends beyond the F-Series as well, with the 2025 Expedition getting a redesigned suspension influenced by the F-150. The Expedition may be one of Ford’s dinosaurs, but its new setup does include modified shocks, springs, and other components. These improvements function to match the Expedition’s combination of passenger comfort, towing capability, and off-road performance. This gives drivers the handling and towing ability they expect from a Ford vehicle.
Fall only took up the job three months ago in April 2026
The departure is likely to once again shake up the administration’s AI strategy
The head of the Trump administration’s Center for AI Standards and Innovation (CAISI) has resigned from his role, Axios has reported.
Chris Fall served in the role for just three months following his appointment in April 2026, and no reason has so far been given for his departure from the agency.
Arvind Raman, the current head of the Commerce Department office that oversees the AI testing institute is set to temporarily take up the mantle as head of CAISI until a permanent alternative is found.
Latest Videos From
Another Trump admin departure
Fall’s departure is just one in a string of Trump administration departures so far in 2026. Before resigning, Fall was responsible for working closely with leading US AI labs such as Anthropic, Google‘s DeepMind, OpenAI, Microsoft and Elon Musk’s xAI to test their frontier and unreleased models for vulnerabilities to prevent them from being maliciously abused.
Advertisement
Trump has placed a big focus on developing AI, with the technology now a key part of both the US economy and its security, with AI models being deployed across federal agencies, police, and the armed forces.
Despite interventions by the US, China has been rapidly closing the gap between its own models and those of US companies. Many Chinese models are cheaper than their US alternatives, making them attractive to US companies feeling the token-cost of US models.
The resignation will likely once again shake up the Trump administration’s AI strategy. The administration has taken a very hands on approach to AI technologies, requiring contractors to provide unrestricted access to their systems and allow the use of their technologies for any lawful purpose. When Anthropic rejected these demands, the company was designated a supply chain risk.
Advertisement
Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed!
Trump has also called on AI companies to provide stakes in their companies in what Trump has said will “create almost a partnership with the American public”. Trump hasn’t revealed exactly how this partnership would benefit the American public.
Physical media collectors have three very different 4K restorations to keep on their radar: Michael Mann’s Manhunter, Hammer’s Dracula, and Criterion’s upcoming 4K edition of 12 Angry Men.
This is a weird trio, which is exactly why it works for me.
Manhunter is Michael Mann before everyone decided they had discovered Michael Mann. It is cold, controlled, stylish as hell, and still one of the most unsettling serial-killer films of the 1980s. It needs 4K because the color, lighting, architecture, and dead-eyed calm are the movie.
Dracula is Hammer at full strength: Christopher Lee, Peter Cushing, blood, capes, religious dread, and enough Gothic atmosphere to make modern horror look like it was assembled in a conference room. If the restored footage and color work are handled properly, this could be the Hammer release fans have been waiting for.
Advertisement
12 Angry Men is the outlier, but only on the surface. No monsters. No synth score. No blood. Just Henry Fonda, twelve men in a room, and a film that gets tighter and more suffocating by the minute. That is exactly the kind of black-and-white classic that can benefit from a serious 4K restoration because every face, bead of sweat, cigarette, shirt collar, and hard stare is what made this film interesting in the first place.
The important caveat: only 12 Angry Men currently has a fully confirmed 4K UHD release date, price, disc configuration, HDR format, audio format, and supplement list. Manhunter and Dracula are confirmed new 4K restorations with theatrical and home entertainment plans, but final 4K UHD disc specifications and pricing have not been fully announced yet.
Related Reviews:
Mann’s Neon Nightmare Needs the Right 4K Treatment
Michael Mann’s Manhunter is returning as Manhunter: The Final Cut, with StudioCanal celebrating the film’s 40th anniversary through a new 4K restoration. The film opens in select U.S. theaters through Rialto Pictures on July 24, 2026, followed by a StudioCanal UK theatrical release on September 25, 2026, with further international releases planned for the fall. 4K Blu-ray and Blu-ray release details are expected later this year.
Advertisement
That means buyers should not treat this as a finalized UHD preorder yet. There is no confirmed U.S. 4K disc price, HDR format, audio format, region coding, packaging, or supplement list at this stage.
What we do know is enough to make this one worth watching. Manhunter was shot by Dante Spinotti and remains one of Mann’s most visually controlled films. Its cold architecture, blues, greens, whites, glass surfaces, and nightmarish calm are central to the experience. A strong 4K restoration should preserve the film’s hard visual edge without scrubbing away the texture.
The cast also makes it more than a Mann completist title. William Petersen, Joan Allen, Dennis Farina, Tom Noonan, and Brian Cox give the film its unnerving edge. Cox’s Hannibal Lecktor, spelled that way in Manhunter rather than Thomas Harris’s original “Lecter,” is not the grand theatrical monster Anthony Hopkins later created. Hopkins is the more iconic nightmare. Cox is colder, quieter, and more clinical, which makes his version unsettling in a very different way.
Hammer’s Full Restoration Is the Collector Hook
Advertisement
Hammer’s Dracula from 1958, released in the U.S. as Horror of Dracula, is also receiving a major 4K restoration. Hammer has confirmed that Peter Cushing and Christopher Lee will return in 4K, with the film coming back to cinemas and making its worldwide home entertainment debut in 4K.
Advertisement. Scroll to continue reading.
This is not just another catalog cleanup. Hammer says the new restoration uses film materials supplied by Warner Bros. and restores the full version of the film as director Terence Fisher intended it before censorship cuts. Hammer also says the release will include the newly restored full-length version alongside the UK theatrical version and the U.S. Horror of Dracula version.
That is the real hook for collectors. Dracula has existed in several versions, and earlier releases have had to deal with incomplete or compromised Japanese elements. Hammer says Warner’s archive uncovered pristine reels of the Japanese version, allowing the restored material to be reintegrated properly.
Advertisement
A combo pack release of the 4K restoration is scheduled to arrive later this year, with theatrical screenings planned for early October in association with Warner Bros.
Pricing has not been confirmed. Final disc specs, HDR format, audio format, region coding, and complete extras have not been announced. Hammer has said this will be part of its Limited Collector’s Edition range, so expect collector packaging and supplements, but do not treat any price as final until Hammer posts the actual preorder.
Why should buyers care? Because Hammer horror lives or dies by color, contrast, and atmosphere. Jack Asher’s cinematography, the blood, the costumes, the candlelit interiors, and Lee’s physical menace all need careful handling. Bad compression and weak grading make Hammer look cheap.
Criterion Sends 12 Angry Men Back to the Jury Room in 4K
Criterion’s 12 Angry Men is the safest confirmed buy of the three because the details are already listed.
The 4K UHD plus Blu-ray special edition arrives September 8, 2026 for $49.95, along with standard Blu-ray at $39.95, and DVD version at $29.95.
The 4K edition includes a new 4K digital restoration, Dolby Vision HDR, and an uncompressed monaural soundtrack. The package includes one 4K UHD disc with the film and one Blu-ray with the film and special features.
Criterion also lists the 1954 television version directed by Franklin J. Schaffner for Studio One, archival Sidney Lumet interviews, production history material, interviews related to Reginald Rose and Boris Kaufman, Lumet and Rose’s 1956 teleplay Tragedy in a Temporary Town, the trailer, English SDH subtitles, and an essay by Thane Rosenbaum.
Technically, this is not a spectacle disc. It is black and white, 1.66:1, mono, and 96 minutes long. A good 4K restoration should improve texture, facial detail, sweat, cigarette smoke, white-shirt contrast, window light, and the claustrophobic feel of the jury room. If the transfer is done right, the walls should feel like they are moving inward.
Apple says it fixed a vulnerability that could expose real addresses behind Hide My Email on or around July 3, but we reproduced the flaw two weeks after the company’s claimed repair date.
The company told 404 Media that it deployed a patch on July 3, 2026, and fully resolved the vulnerability. AppleInsider successfully reproduced the behavior on July 17.
Our test found that a sender who possessed a specific Hide My Email alias could reveal the real address behind it with much technical knowledge.
We’ve reached out to Apple about the matter. We will update when we hear back from the company on the matter.
Advertisement
Apple spent roughly a year responding to the flaw
The vulnerability affected iCloud Hide My Email, which creates aliases that forward messages to a user’s real inbox. Users can provide a different alias to each website or service instead of sharing a permanent email address.
EasyOptOuts co-founder Tyler Murphy began reporting Hide My Email vulnerabilities to Apple in mid-2025. 404 Media says Murphy first alerted Apple in June 2025, while EasyOptOuts’ published timeline lists a related report on July 9.
Murphy and EasyOptOuts co-founder Ben Weiner exchanged reports and test results with Apple for roughly a year. EasyOptOuts says Apple declared the vulnerabilities fixed on March 3 and June 30, but the researchers reproduced the problems after both claims.
Murphy later contacted 404 Media, which reported the flaw on July 1 without publishing instructions that could help others exploit it.
Advertisement
Apple told 404 Media that a patch deployed on July 3 fully resolved the vulnerability. The company hasn’t said whether July 3 marked the start or completion of the deployment, but we expect it is the former.
AppleInsider reproduced the flaw on July 17
AppleInsidertested the flaw on July 17 and confirmed that the process could reveal the real address behind a Hide My Email alias.
The process required little technical knowledge once the sender possessed a specific alias, though each address had to be targeted individually. We withheld the instructions because publishing them while the vulnerability remained reproducible would have created unnecessary risk.
The test doesn’t prove the vulnerability remained active for every user or mail provider. It does show that July 3 can’t be treated as a definitive endpoint without further explanation from Apple.
Advertisement
Apple hasn’t explained whether the patch was deployed in stages, when the rollout finished or why the July 17 test succeeded. Without those details, the company’s claim of a complete July 3 resolution remains difficult to reconcile with independent testing.
The vulnerability was real, but its practical risk was limited
The flaw undermined the core promise of Hide My Email. A sender who possessed an alias could potentially discover the permanent address that the service was supposed to conceal.
EasyOptOuts said every address in its limited volunteer tests was vulnerable. Murphy and Weiner also said the vulnerability had been fixed, but warned that the privacy risk may continue after the patch.
Hide My Email on an iPhone
Murphy and Weiner told 404 Media that rejected messages may have exposed real addresses still retained in third-party logs. EasyOptOuts recommends treating addresses linked to aliases created before July 7 as potentially exposed.
EasyOptOuts’ warning doesn’t mean every Hide My Email user was affected. A sender first needed a specific alias, which limited the number of accounts that person could target, and there’s no public evidence of a coordinated exploitation campaign.
Advertisement
The vulnerability didn’t expose passwords, unlock Apple Accounts or grant access to inboxes. The demonstrated consequence was the loss of an alias’s privacy protection, which could allow a permanent address to be connected with leaked or publicly available information.
A proposed class action filed after the disclosure seeks repayment of subscription fees and other relief. The complaint doesn’t allege that the named plaintiff’s address was exposed or that anyone exploited the vulnerability against him.
Apple still needs to explain the conflicting dates
The available evidence supports calling the flaw a genuine privacy failure. It doesn’t show widespread exploitation or measurable harm to subscribers.
Apple says the vulnerability has been eliminated, and Murphy and Weiner also say the bug has been fixed. We can’t reproduce it today, after our July 17 testing.
Advertisement
The available evidence doesn’t show that users need to disable Hide My Email entirely. Users should nevertheless treat real addresses attached to older aliases as potentially disclosed and understand that deleting an alias won’t erase information already retained in another provider’s logs.
.jpg)
You must be logged in to post a comment Login