Published file metadata, source observations, indicators, and research history in one bounded catalog.
2,400+Reachable catalog records
83 / 100Current catalog page
24Maximum records per page
0Samples downloaded by this page
Feed collection status
Sources represented on this page. Publication and observation dates describe the content; successful batch dates describe local collection. A recent batch does not certify complete coverage or current provider availability.
intelligence_src_33e452677074d4c3bdRecent successful batchLast successful batch: Last attempt:
intelligence_src_350d40f2cc69277bcaRecent successful batchLast successful batch: Last attempt:
Stale means older than twice the recorded collection interval, with a minimum of one hour; without an interval, the threshold is 24 hours. Cached content remains available while a feed is stale, paused, or unverified.
2,400+ MATCHING RECORDSMETADATA ONLY / SOURCE ATTRIBUTION PRESERVED
Summary There is a vulnerability in the Huawei Kirin SoC’s DDR Controller (DMSS) Access Permission system which allows the Linux kernel to bypass memory access restrictions and directly compromise multiple privileged subsystems of the SoC. As demonstrated by CVE-2021-3710, CVE-2021-39991, CVE-2021-37115, and CVE-2021-39986, read and write of critical system memory, including secure memory regions, is possible via those subsystems. Therefore, this vulnerability combined with one of CVE-2021-3710, CVE-2021-39991, CVE-2021-37115, or CVE-2021-39986 results in a fully realized chain of elevation of privileges from a kernel-level write primitive to total control of the secure world (TEE). The vulnerability was fixed in February 2022.
First reported May 18, 2022 00:00 UTCCatalog metadata updated Oct 13, 2026 08:35 UTCOpen evidence record
Summary In a previous advisory we disclosed multiple vulnerabilities within the NPU device’s mmap handler and discussed how it exposes sensitive kernel data. This advisory focuses on the implementation errors in the same handler. The mapping function ignores the requested size parameter and fails to back regions of the mapped virtual memory with an actual mapping. Trying to access these holes or requesting an mmap that is smaller than the memory mapped by the kernel driver results in a kernel crash, most likely caused by invalid page accesses. The mmap handler is exposed through the /dev/davinci0 character device. Due to the applied selinux policy, access to this device is restricted to the hiaiserver system process.
First reported Aug 02, 2021 00:00 UTCCatalog metadata updated Oct 13, 2026 08:35 UTCOpen evidence record
In this advisory we are disclosing a vulnerability in the Huawei log device that allows any unprivileged process to disclose sensitive information from the kernel. Huawei kernels are shipped with custom log devices (/dev/hwlog_dubai, /dev/hwlog_exception and /dev/hwlog_jank) that facilitate better system diagnostics through a series of ioctl calls. One of these diagnostics module is referred to as zrhung, and it provides information and configuration options to monitor hung processes. The implementation of the config set ioctl contains a race condition vulnerability that allows an attacker to free the underlying buffer that holds the configuration data. Consecutive config get ioctl calls use the dangling pointers to read and disclose, potentially sensitive, dynamically allocated kernel data.
First reported Jul 24, 2022 00:00 UTCCatalog metadata updated Oct 13, 2026 08:35 UTCOpen evidence record
Once upon a time, I started at TASZK Security Labs as an intern. My internship project was about hacking hypervisors, and the target we picked for it was Huawei’s HEE (Hypervisor Execution Environment). The research was carried out in late 2020 to early 2021. Although we didn’t publish this work all the way until 2025, it was kept relevant by the fact that the issues remained unaddressed, to the best of our knowledge. If that piqued your interest in the story of the disclosure, please check out our recent Troopers/LeHack talk titled Eastern Promises. In this blogpost, I’ll focus on the technical aspects of the project.
First reported Sep 24, 2025 00:00 UTCCatalog metadata updated Oct 13, 2026 08:35 UTCOpen evidence record
Summary The NPU device’s kernel driver implements a set of ioctl handlers one of which uses unsanitized user data as an offset to retrieve a kernel structure. Fields of the structure are written with user provided values. A malicious actor can use this vulnerability to overwrite kernel memory with controlled data. The mmap handler is exposed through the /dev/davinci0 character device. Due to the applied selinux policy, access to this device is restricted to the hiaiserver system process. Because of these limitations a practical attack would need to target the hiaiserver first. The /dev/davinci0 device exposes a series of ioctl handlers, one of which is devdrv_ioctl_load_stream_buff.
First reported Aug 02, 2021 00:00 UTCCatalog metadata updated Oct 13, 2026 08:35 UTCOpen evidence record
Remember OpenAI’s video model Sora and all the great Sam Altman memes that came from it? Earlier this year, the ChatGPT-maker canned both the model and the TikTok-like app it powered as part of broader culling of server-intensive “side projects” even as rivals xAI and Google forged ahead with their own video models.Well, there’s a chance that a version of Sora is coming back from the dead.A stealthy startup developing AI video for filmmakers, which we revealed last month, is in talks with OpenAI to access Sora, tweak it and sell software or other services using it, according to a person with knowledge of the talks. Hollywood mogul Jeffrey Katzenberg, former Sora research lead Bill Peebles, and former Dropbox CFO Sujay Jaswa co-founded the startup in recent months. (We don’t know the startup’s name).
First reported Oct 12, 2026 14:00 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
To understand the decision-making of AI, developers often look at the text that models produce as they reason through problems. But these “chains of thought” are increasingly hard to read, which is prompting AI developers to consider other approaches. One alternative, known as “mechanistic interpretability,” seeks to explain the inner workings of AI models, rather than just relying on their outward behavior. Leading AI companies have formed mechanistic interpretability teams in recent years, and the field has spawned several startups, the most prominent of which is Goodfire AI, whose chief scientist Tom McGrath joined the latest episode of AI Deep Dive. (The full episode is available here.)
First reported Oct 13, 2026 00:43 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
Meta Platforms has banned advertisements from ByteDance on its platforms in the U.S. and several other markets, Bloomberg and other media outlets reported. Meta implemented the ByteDance ad ban in the U.S., Canada, Egypt, Indonesia, Japan, Thailand and Vietnam, according to Bloomberg. The ban ...
First reported Oct 12, 2026 10:43 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
Meta Platforms unveiled two new AI models, Muse Wonder and Muse Wonder Full, on Monday as it seeks to strengthen its position in the competitive AI race. CEO Mark Zuckerberg announced the models on Threads, describing them as significantly larger and more advanced than the company’s most recent ...
First reported Oct 12, 2026 17:31 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
• The Big Read: Meet Zach Frankel, the secretive investor behind Cognition, Ramp• Why techies are hoarding AI compute in their homes• Plus, Recommendations—our weekly pop culture picks: “Bram Stoker’s Dracula,” “The AGI Chronicles” and “Your Mother Your Mother Your Mother”Mark Zuckerberg, the founder and CEO of Meta Platforms, was only six years old in 1990, the year the Human Genome Project got launched.That project, a multibillion-dollar international effort led by the National Institutes of Health and the Department of Energy, resulted in the announcement 10 years later of an awe-inspiring milestone: the first sequencing of the human genetic code—the building blocks of life. When President Bill Clinton described the achievement in 2000, he framed it in grand terms: “We are learning the language in which God created life.”So it was intriguing to hear another scientific collaboration compared to that earlier achievement at a reception this week hosted by Biohub, the primary philanthropic effort of Zuckerberg and his wife, Priscilla Chan. The event this week is one of a number of recent moments that herald a changing of the scientific guard from major universities to big tech companies. Speaking at the Treehouse Hotel in Sunnyvale, Calif., Zuckerberg told the gathering of around 100 tech executives, scientists and academics that he and Chan had started the first version of what is now called Biohub almost 10 years ago to the day with the goal of trying to cure and prevent all diseases by the end of the century. “We got a lot of eye rolls,” Zuckerberg said. “People thought that was kind of crazy.” Thanks to AI and other advances in technology, he added, “That no longer seems quite so crazy.”He then turned to Lori Goler, president of the Chan Zuckerberg Initiative, who announced that Biohub is partnering with the National Institutes of Health and the Department of Energy in a $1.8 billion investment to create new technologies and build the scientific infrastructure required to drive “AI-powered biology.” Biohub, which launched what it calls the “Virtual Biology Initiative” in April with a $500 million commitment, also said at the event that Meta, Google DeepMind and Isomorphic Labs, a company spun off of DeepMind, have agreed to kick in $300 million.“We were inspired by the Human Genome Project,” Goler told the group. Goler added that they hope the initiative “will have the same level of impact and will be as important as the Human Genome Project has been.” (Anyone going to The Information’s WTF Summit later this month will hear more from Goler on the future of AI biology.)Despite the evening’s focus on collaboration, science can be an isolating discipline in its daily practice. When I visit labs, I am always amazed by how quiet they are. The scientists are usually intently concentrating on not making a false step that could lead to the contamination of an experiment. And yet, once the results leave the lab and reach the public, clamor often ensues. Controversial findings spark debate over ethical and moral questions. Transformative discoveries are celebrated with an element of pageantry and recognition. Both are reminders of our common humanity—and that the ultimate goal of science is to make all our lives better. This leads to invitations to the White House like the one that happened back in 2000 with the scientists at the helm of the Human Genome Project. Or even more personal celebratory videos like the ones posted on social media all this week by happy scientists who received early morning calls from Sweden learning they had won Nobel Prizes. (How can you not be moved by this video of the Stanford neuroscientist Karl Deisseroth telling his kids he won the Nobel Prize for medicine?)Zuckerberg is one of the most prominent symbols of how AI is altering the scientific landscape and bringing new leaders into an arena once dominated by major universitie
First reported Oct 10, 2026 15:01 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
Imagine Epic Games, Apple and DoorDash were developing their own large language models now from scratch to compete against the likes of OpenAI and Anthropic. That’s basically what’s happening in China, where a trio of big names from other parts of the tech industry have jumped into an already-crowded race to make cutting-edge AI. MiHoYo, a Shanghai-based videogame maker behind such hits like Genshin Impact, is developing its own LLMs with the goal of becoming a top-tier lab in China in the next couple of years. Xiaomi, a leading smartphone maker that also sells electric cars, released its latest LLM two weeks ago. And Meituan, China’s biggest food delivery and local services app, in June unveiled a more powerful version of the LLM it launched last year.
First reported Oct 09, 2026 13:01 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
The basement of Gary Flake’s contemporary home in a quiet neighborhood in Bellevue, Wash., is furnished with everything you’d expect in a multipurpose man cave. There’s a cozy-looking couch situated between a large screen and a digital projector, a Peloton treadmill and a rowing machine. The real showpiece of the space, however, is a custom computer that resembles an oversized erector set, one with seven large Nvidia graphics processing units dangling from its frame. Flake, a former Microsoft executive and startup founder, built the computer to train custom AI models on terabytes of data from electroencephalograms, the tests that measure electrical activity in the brain—part of a startup project he has since moved on from. To train his AI models, Flake could have paid a cloud service to train them on the industrial-strength GPUs inside their data centers, but he said it was far cheaper to assemble his own machine out of discounted parts (total cost: under $6,000) than to fork over fees to Amazon Web Services or Microsoft every month. That’s true even when including the $10,000 to $20,000 he spent for electrical upgrades and a new HVAC system he bought to keep his basement cool while his AI computer throws off a lot of heat. “I want to own,” said Flake, who plans to begin using the computer again for a new AI project in the coming months. “I don’t want to rent.”
First reported Oct 10, 2026 14:00 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
Why is Anthropic driving its partners to speed up the $530 billion of data center leasing deals it has signed? It’s simple: Every server Anthropic plugs in sooner to run its products could be wildly profitable, according to fresh analysis of upcoming AI data center sites, including one Anthropic will rent outside Baton Rouge, La.
First reported Oct 11, 2026 15:01 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
Bain Capital-owned Bridge Data Centres is in talks with banks to increase the size of a loan from $2.8 billion to about $6.6 billion for its Malaysia operations, Bloomberg reported. The raise would be one of the largest debt financing deals in Asia’s data centre industry. Bridge Data Centres, ...
First reported Oct 12, 2026 10:24 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
In the summer of 2025, Microsoft was fielding customer complaints that its AI assistant, Copilot, was buggy and not good enough at automating work in Excel or PowerPoint. In response, Charles Lamanna, then a Microsoft president and rising star within the giant firm, went to CEO Satya Nadella with a bold proposal: Microsoft needed to consider paying for AI models from Anthropic to make Copilot work better. “At the time, that was heresy,” Lamanna said in an interview.
First reported Oct 12, 2026 13:01 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
Anthropic said Friday that is AI used the internet to inadvertently fill out government forms with incomplete information and also submitted a false homicide crime tip to the Philadelphia police department. Anthropic said police flagged the homicide tip as spam, and Axios reported that Anthropic ...
First reported Oct 10, 2026 06:00 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
Dating app company Bumble is in advanced talks to be taken private by a holding company founded recently by former Sequoia’s partner Ravi Gupta, the Wall Street Journal reported Monday. A deal would value Bumble at roughly $1 billion, a steep premium to its last market cap of around $300 million ...
First reported Oct 12, 2026 19:03 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
As startups scramble to get access to AI servers, more venture capital firms are finding ways to secure compute on their behalf. Andreessen Horowitz, which two years ago set up a cluster of 20,000 graphics processing units for its startups, is in talks to expand its offering, potentially via two data center companies it has backed, according to people familiar with its plans. Meanwhile, General Catalyst is building a data center to supply compute for its startups, said two other people. And several other VC funds have used brokers to find sought-after chips for their companies.
First reported Oct 12, 2026 17:42 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
Microsoft is overhauling the leadership of its Azure cloud business, cloud chief Scott Guthrie told staff Monday, as the company consolidate units and reorients itself around AI. As part of the reorg, Arun Ulag, who previously oversaw Azure’s database products, will now also oversee ...
First reported Oct 13, 2026 00:59 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
Microsoft on Monday announced the latest generation of its homegrown AI chips, dubbed the Maia 300, and said it plans to set up “multi-gigawatt deployments” of the chip starting next year. The move shows how Microsoft is aiming to reduce its dependence on Nvidia-designed GPUs, which currently ...
First reported Oct 12, 2026 13:47 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
If it’s Monday, it must be time for another AI model release. This time, it was Meta Platforms’ turn, unveiling Muse Wonder, which it claims can “deliver state-of-the-art performance” on certain tasks. It was a triumphant moment for the company, whose AI efforts previously went through a slump that prompted a massive hiring binge over the past 14 months or so. It now claims it is “pushing the frontier” of AI.That would be notable if only everyone wasn’t claiming to be doing the same thing. Two weeks ago, Google unveiled its latest model, Gemini 4 Argon, which it says “delivers frontier performance in complex workflows” across various subjects (you wouldn’t want it to deliver frontier performance in simple workflows, would you?). In early September, OpenAI announced GPT-6 Astra, which it said was “state of the art on computer use, browsing, software engineering, cybersecurity, science and professional work.” Also in September, Anthropic introduced Claude Fable 5.1 and Mythos 5.1, which it claimed to be “the world’s most advanced models for coding and knowledge work.”
First reported Oct 13, 2026 00:00 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
SoftBank Group CEO Masayoshi Son is looking to raise up to $100 billion from Gulf investors to finance his AI investments, The Financial Times reported. Son has held fundraising discussions in recent weeks in the United Arab Emirates and other parts of the region, according to the FT. He plans ...
First reported Oct 09, 2026 09:26 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
The small agency within the Commerce Department tasked with testing new AI models appears to be getting a refresh. In September, the Trump administration began the process of restructuring the Center for Advancing Innovation and Standards for Super Intelligence so that three of its senior officials report directly to the White House, according to a White House official.
First reported Oct 12, 2026 18:36 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
Nuclear reactor startup Valar Atomics says it has sued one of its early backers, Day One Ventures, in Delaware court after shutting out the venture firm from its recent $1 billion funding round. Founders rarely sue their own investors, but Valar is asking the court to declare that the startup ...
First reported Oct 10, 2026 00:58 UTCCatalog metadata updated Oct 13, 2026 08:33 UTCOpen evidence record
The public shared-host view reaches the newest 2,400 matching records. Narrow the filters or search an exact MD5, SHA-1, or SHA-256 value to reach a specific record.
A filename, label, or report mention alone does not establish maliciousness. Observation dates describe cited sources; collection dates describe this local metadata catalog. No malware samples are stored, fetched, or executed by these pages.