Microsoft on Monday launched its first cybersecurity-specialized model alongside a new AI cybersecurity platform at a small event in San Francisco, taking a big swipe at major players in the space — namely Anthropic, Google and OpenAI.

The company describes MAI-Cyber-1-Flash as a model that’s built “to find challenging vulnerabilities in complex codebases.” The model is built to animate MDASH, Microsoft’s harness dedicated to software vulnerability identification and remediation.

The new security platform is dubbed Perception, and it’s designed to deploy teams of agents to assist with and automate various security workflows, including identifying and remediating bugs. The platform can also integrate with MDASH.

The company claims MAI-Cyber-1-Flash is significantly more powerful (and more cost-effective) than competitor models, based on its performance on an established AI cybersecurity benchmark.

“We’re very very excited to announce our results,” said Mustafa Suleyman, the co-founder of DeepMind and current CEO of Microsoft AI. “We have MAI-1 Cyber Flash binded [sic] with GPT 5.4 inside of the MDASH harness — which beats out Gemini, GPT 5.5 Cyber, GPT 5.6 Sol, and Mythos 5 on Cyber Gym, which is the primary benchmark that we all use. The golden benchmark.”

“We’re shipping this into production immediately,” he added.

Noting that hackers are increasingly using AI in their cyberattacks, Hayete Gallot, Microsoft’s vice president for security, described Perception as a way for enterprise defenders to “defend against AI with AI at the scale and speed that the attackers have.”

Perception uses agentic red teams, blue teams, and green teams. The red teams can provide detailed simulations of potential attacks — providing context about potential threat actors and the likely vulnerabilities that they might exploit. Blue teams are dedicated to detecting and triaging existing bugs, while green teams take “corrective actions” against those bugs.

Dave Weston, the lead engineer for Perception, described the platform as a massive efficiency upgrade for corporate defenders. “We’ve gone from this taking hours and hours of manual work from multiple specialized folks across the security organization — appsec hunters, remediation engineers, you name it — and in minutes, we have a fix for all of this. Not only do we discover the issues and prioritize them, but we have detection, posture fixing, and even a code fix.”

Though AI has offered new defensive capabilities to companies, its availability to cybercriminals has given rise to a dazzling array of potential threats.

Microsoft’s new security tools, which the company said will be available in preview on November 3, will enter an increasingly crowded field of AI cybersecurity solutions. Earlier this year, Anthropic launched Mythos, a security platform that was released to a small coterie of partner organizations through a program called Glasswing. OpenAI has also launched its own security solution in May through a program called Day Break.

Apple is being sued by three people who claim approximately $1.8 million in Bitcoin was stolen after downloading and using a fraudulent Sparrow Wallet application from the App Store.

The complaint, filed on July 24 in California, alleges that Apple failed to adequately review and monitor applications distributed through the App Store while promoting the marketplace as a safe and trusted source for software.

Plaintiffs James Ramirez, Christopher Ellis, and Jalen Delgado say the malicious application impersonated the legitimate Sparrow Bitcoin wallet and instructed them to enter their seed phrases.

After entering those secret recovery credentials, the victims allegedly had their Bitcoin transferred to cryptocurrency wallets controlled by the scammers.

The legitimate Sparrow Wallet is a desktop application available for Windows, macOS, and Linux, and does not offer an iOS version.

However, Sparrow Wallet’s developer says that scammers have repeatedly published applications in Apple’s App Store that impersonate the cryptocurrency wallet.

Victims lost Bitcoin after earlier warnings

According to the complaint, Delgado downloaded the fraudulent application on or around May 1, 2025, entered his seed phrase, and shortly afterward discovered that 1.05033242 Bitcoin, worth approximately $120,000, had been transferred to a scammer.

Ramirez allegedly downloaded the app on July 25, 2025, and lost 7.4 Bitcoin, valued at approximately $875,000. The complaint says Ramirez reported the fraudulent app and theft to Apple that same day, but alleges Apple never contacted him regarding that report or subsequent reports.

Just over a week later, on or around August 3, Ellis allegedly also installed the Sparrow app from the App Store. After entering his seed phrase, Ellis discovered that approximately $840,000 in cryptocurrency had been transferred to a scammer and immediately reported the app and theft to Apple.

The complaint also claims that Apple ranked the fraudulent Sparrow app and included it in curated cryptocurrency app collections, effectively recommending it alongside legitimate applications.

However, Apple had allegedly been warned about fraudulent Sparrow Wallet apps more than a year before the plaintiffs’ losses.

The complaint includes a January 6, 2024 post from Craig Raw, the developer of the legitimate Sparrow Wallet application, who said a scam Sparrow Wallet app remained available in the App Store despite him and others reporting it weeks earlier.

Raw also warned users to only obtain Sparrow Wallet through its official website rather than trusting an application solely because it was available through an app store.

MacRumors also shared a more recent attempt by Raw to protect users from the fake Sparrow applications on the Apple App Store by submitting an unpublished placeholder app that explained that mobile apps using the “Sparrow Wallet” name were fake.

According to Raw, Apple initially flagged his developer account for termination over alleged dishonest activity, but later reversed the decision.

Apple told BleepingComputer that it acted quickly to remove applications impersonating Sparrow Wallet and terminated the developer accounts associated with them.

Apple also said developers who believe content distributed through its services infringes their intellectual property can submit a dispute to the company’s legal department.

Customers can separately report suspected App Store scams and fraud through Apple’s Report a Problem service, and Apple said it takes immediate action when applications are found to violate its guidelines. 

The plaintiffs are seeking reimbursement for the stolen cryptocurrency, compensatory and punitive damages, restitution, attorneys’ fees, and other damages.

They are also asking the court to require Apple to improve and publicly disclose its procedures for detecting and removing fraudulent applications, while also introducing warnings about the risks associated with cryptocurrency apps.

Security teams log 54% of successful attacks and alert on just 14%. The rest move through your environment unseen.

The Picus whitepaper shows how breach and attack simulation tests your SIEM and EDR rules so threats stop slipping by detection.

Advertisement

Get the whitepaper

What just happened? Chinese memory maker CXMT is now the country’s most valuable company after an impressive first day on the market that exceeded expectations. The stock’s initial public offering price was set at CNY 8.66 but ended up closing at CNY 49.00 – a gain of nearly 466%.

CXMT, short for ChangXin Memory Technologies, is one of several major players in the memory market that are profiting from the ongoing AI boom. At present, CXMT ranks fourth overall with an eight percent share behind Samsung, SK Hynix, and Micron with shares of 36%, 29%, and 24%, respectively.

Larry Yang, chief economist of First Seafront Fund Management, told AFP that CXMT’s first-day performance reflects investors’ overwhelmingly bullish sentiment regarding the company.

The memory maker closed the successful day of trading with a market cap of around 3.65 trillion yuan, or roughly $540 billion. Fortune said it was mainland China’s second largest IPO ever, following that of Agricultural Bank of China back in 2010.

Yang added that the newly raised money will give CXMT the ability to expand production capacity, in addition to boosting its global influence. Being able to tap a homegrown hardware maker should also benefit China in its AI race against the US.

CXMT was founded in 2016, just in time to leverage the pandemic-fueled PC buying surge. Then came the AI boom and as you can imagine, hardware makers and investors have been singing all the way to the bank. Consumers, meanwhile, have been met with soaring prices on smartphones, game consoles, and virtually everything else that relies on a memory chip to operate.

Worse yet, there appears to be no relief in sight. Earlier this month, SK Hynix said it expects the memory crunch to peak in 2027 and continue through at least 2030.

According to AFP, Apple is now testing DRAM chips from CXMT for use in future products. Signing on as a supplier for the Cupertino-based company is no doubt high on CXMT’s list of short-term goals.

What just happened? Samsung Electronics has expanded its semiconductor partnership with Broadcom to focus on AI chips, bringing together chip design, manufacturing, and memory in a deal that could run through the end of the decade. The companies said the agreement covers memory chips, contract manufacturing, and advanced packaging, with total activity expected to exceed $200 billion by 2030. The partnership gives Broadcom deeper access to Samsung’s production capabilities, while Samsung secures longer-term demand for its most advanced fabs.

At a technical level, the deal is built around closer integration between Broadcom’s ASIC designs and Samsung’s manufacturing stack. Broadcom’s next-generation communications chips will be produced using Samsung’s sub-2-nanometer process, putting them at the leading edge of chip scaling, where efficiency and heat management are becoming increasingly challenging.

The two companies are also working on next-generation high-bandwidth memory, a critical component for AI systems that need to move large amounts of data quickly. As models grow larger, memory bandwidth has become a limiting factor, making tighter coordination between logic chips and memory increasingly important.

The partnership reflects a broader shift in how AI hardware is being developed. Large tech companies are moving toward custom-built chips designed for specific workloads rather than relying solely on general-purpose GPUs. That trend is pushing chip designers and manufacturers toward closer, longer-term collaborations.

For Samsung, the deal is also about strengthening its position in the foundry market, where it competes with TSMC. Securing consistent, high-volume work from a company like Broadcom could help improve utilization at its advanced manufacturing facilities.

Samsung has been trying to attract more customers for its most advanced nodes. Earlier this year, the company said it expected to win additional 2-nanometer orders following discussions with major tech firms. Its leadership has also pointed to ongoing talks around next-generation memory, including HBM4E and HBM5.

Samsung has landed large contracts before, including a $16.5 billion deal to produce chips for Tesla. But the Broadcom agreement stands out for its scope, covering multiple parts of the semiconductor process rather than a single product line.

The timing comes as demand for AI infrastructure continues to grow and become more specialized. Performance gains are increasingly tied to how chips, memory, and packaging work together rather than improvements in any single component.

ai and ml

The Open Security AI Alliance says the Hugging Face/OpenAI mess proves frontier labs can’t be trusted to properly secure sensitive systems

In the wake of OpenAI agents attacking Hugging Face, Nvidia has recruited a new posse of partners to promote open source models as the security solution the industry needs. The AI arms dealer announced the foundation, the Open Secure AI Alliance, in a blog post today, describing the mission of the group being “to ensure defenders everywhere have open, frontier tools they can trust and control.”

Partners in the group are numerous, ranging from established tech giants like Microsoft, Red Hat, HPE, IBM, and Adobe to newer groups like Palantir, SpacexAI, Hugging Face, and The Linux Foundation. What all the founding members have in common, Nvidia said, is that they agree open source AI models are a fundamental part of modern cybersecurity, just like prior open source tech has been for the infosec space. 

“The United States and its partners now face a choice in AI security: whether the defenses that protect our infrastructure will sit inside a few opaque systems or be built on open models, harnesses and tools that any defender can study, adapt and deploy,” Nvidia said in the announcement. 

The claims in many ways echo the pleadings from tech industry heavyweights made in an open letter to US government regulators last week. That letter, signed by many of the same companies that are part of the founding OSAA cadre, essentially argues that regulators should ensure Anthropic, Google, and OpenAI don’t end up with total control of the US AI market, and that open-weight models should be given a seat at the table, too. 

The new alliance is arguing that, not only do open-weight models need to be allowed to proliferate in the US, but they also need to be considered a fundamental part of the security puzzle.

For those unfamiliar with the Hugging Face incident, a group of autonomous OpenAI agents, operating in a sandbox and stripped of guardrails to test their full capability to solve cybersecurity puzzles, exploited a pair of zero-days to escape and gain access to the internet. For some reason, the bots thought the solution to the problems they were posed could be found in Hugging Face systems, so they broke in and accessed a bunch of private information and hijacked some credentials. 

When Hugging Face turned to closed-source US frontier AI lab bots to examine the incident and help figure out what happened, those tools declined to help because they thought the data Hugging Face was trying to examine was itself malicious. Hugging Face turned to Chinese-made GLM 5.2, hosted on its own infrastructure, to figure things out. 

“That incident showed a practical truth,” said Nvidia. “When defenders cannot inspect, adapt and run advanced AI on their own infrastructure, their ability to respond is constrained at exactly the moment speed matters most.”

Only open-source AI models, which China leads development on, can fill that role, the OSAA argues, and it’s prepared to counter those who say open models are a threat: Just look at what happened last week and it’s readily apparent that closed source models are dangerous too. 

The Alliance is pooling its efforts to give security pros access to essential open tools. Nvidia said that it’s participating by releasing its Object-Oriented Agent project on GitHub, HPE is contributing its SPIFFE/SPIRE zero-trust AI identity framework, Hugging Face has handed its Safetensors transparent AI model weight formatting to the PyTorch Foundation, and SpaceXAI has open-sourced Grok Build (though the reason behind that doesn’t appear to be entirely benevolent).

In addition, IBM and Red Hat have released Lightwell, an automated open-source vulnerability remediation platform, while Microsoft has come out with MDASH, a multi-model agentic scanning harness to automate bug discovery and remediation. Those efforts, while not open source themselves, are still a sign that Alliance members “are building an open defense stack,” Nvidia said. 

The OSAA ended its announcement with another call for policymakers not simply ban open-source AI models, as doing so “would weaken defensive capacity and risk concentrating power, dependence, and vulnerability in a few closed providers,” the group said.

Many providers, as we saw last week, are more concerned with protecting themselves than helping victims of autonomous cyber attacks respond quickly. 

Clement Delangue, cofounder and CEO of Hugging Face, said in a post on X that he spoke to OpenAI over the weekend about last week’s incident and asked the company to provide funding to support the development of better open-source AI cyber defenses. It’s not clear if the company plans to fulfill that request; it’s not a founding member of the Nvidia-led OSAA. Neither is Google or Anthropic, for that matter. 

We reached out to all three companies for their take on the new initiative, but didn’t hear back from any of them. ®

Samsung could be preparing its busiest flagship launch schedule yet.

According to a new report from South Korea, the company is planning a “4+4” strategy for 2027. This would see eight premium Galaxy smartphones launch across the year, including a second-generation tri-fold device.

If accurate, the roadmap would significantly expand Samsung’s flagship lineup as competition in the premium smartphone market continues to intensify.

The report claims Samsung will split its launches into two waves. The first half of 2027 is expected to be led by the Galaxy S27, Galaxy S27+, Galaxy S27 Ultra and, for the first time, a Galaxy S27 Pro. Thus, the Galaxy S family will have four flagship models instead of the usual three.

The second half of the year is reportedly reserved for foldables. Alongside the Galaxy Z Fold 9, Galaxy Z Fold 9 Ultra and Galaxy Z Flip 9, Samsung is said to be preparing the Galaxy Z TriFold 2. This would mark the return of its most ambitious foldable design.

Advertisement

According to the report, Samsung has already shared details of the expanded lineup with supply chain partners. Furthermore, all four foldables are expected to launch together around July 2027.

Very little is known about the TriFold 2 itself. Although the report suggests Samsung isn’t planning any major changes to its display. The foldable is expected to retain a 9.96-inch main screen with a 4:3 aspect ratio. Meanwhile, earlier rumours have pointed to a noticeably slimmer chassis with all three sections sharing a more consistent thickness.

Production, however, could see a bigger change. Samsung’s first tri-fold device is expected to be produced in relatively limited quantities. Nevertheless, industry sources believe the company will significantly increase manufacturing for its successor if demand proves strong enough.

Advertisement

The move would also continue Samsung’s steady expansion of its foldable portfolio. What was once a two-device lineup has already grown. This year’s range adds a third foldable alongside the Galaxy Z Fold and Galaxy Z Flip series.

Advertisement

The report suggests the expanded flagship strategy could help the company strengthen its position ahead of Apple’s widely anticipated 20th anniversary iPhone lineup. Apple’s lineup is also expected to arrive in 2027.

For now, the roadmap should be taken with a fairly large pinch of salt, as product plans can change a lot before launch. Still, if the report proves accurate, next year could mark the company’s biggest flagship refresh in years.

Advertisement