Meta will pay up to $18B and cap teen Facebook and Instagram use at two hours daily after nearly all US states sued over child safety.

Meta will pay up to $18 billion over the next decade and impose real usage limits on teenagers using Facebook and Instagram, settling claims that the company deliberately designed its platforms to addict children. The deal ended a federal trial mid-stream, right as Instagram head Adam Mosseri had begun testifying and Mark Zuckerberg was expected to take the stand next.

The timing made sense given the huge potential penalties. Four states, California, Colorado, Kentucky, and New Jersey, were seeking up to $200 billion in damages. Before the trial, Meta said they could demand as much as $1.4 trillion. Against those figures, the $18 billion settlement looks relatively small, although it still equals about three to four months of Meta’s profits.

The most important part of the deal is what Meta agreed to change. Teenagers will be limited to two hours a day on Facebook and Instagram. Meta will also block access between midnight and 6 a.m. unless a parent gives permission, and it will turn off most push notifications during school hours.

“The focus of this case was to protect our kids,” Colorado Attorney General Phil Weiser said in a statement reported by Reuters. “The relief we are getting in this settlement is very meaningful and well beyond what any court has ordered or is likely to order.””

What the settlement leaves unchanged matters too. Meta does not have to stop using personalized recommendations or targeted ads for teenagers. It also does not have to remove specific types of content that researchers have linked to negative effects, such as posts that can make users feel worse about their bodies. A two-hour limit is still a meaningful restriction, but Meta can continue trying to maximize engagement during those two hours.

The deal also creates an interesting financial incentive. Of the roughly $16.7 billion going to 47 states, Washington D.C., Puerto Rico and other territories, about $12.7 billion is guaranteed. The remaining $5 billion depends on whether Snapchat, TikTok and YouTube introduce similar protections for teenagers. This gives Meta a financial reason to push its competitors to adopt the same rules, which is why the company reportedly plans to use newspaper ads to encourage TikTok and YouTube to follow suit.

Separately, Wednesday’s settlement also resolved lingering state privacy claims tied to the Cambridge Analytica scandal, with Meta agreeing to pay $459 million on top of everything else. That’s an old wound getting stitched up alongside a much newer one, in the same afternoon.

Not every state joined the settlement. New Mexico stayed out after winning a $567 million public nuisance ruling against Meta earlier this month, on top of a separate $375 million jury verdict. Attorney General Raul Torrez said the settlement didn’t include some changes his case had pushed for, including stronger protection against adults targeting children and a ban on sexualized AI chatbot interactions with minors. Still, he called the deal a step forward.

Florida rejected the settlement altogether. Attorney General James Uthmeier said the payouts amount to “peanuts” compared with the harm caused and said Florida would take Meta to trial instead.

Legal experts already see the settlement as a possible model for future cases. Northwestern law professor James Speta said Meta and other tech companies faced growing pressure to change anyway, from Congress, state lawmakers and the public. That makes the settlement more than a single case: it could set a standard that courts and regulators use when judging other platforms.

Thousands of similar lawsuits from individuals, school districts and municipalities are still moving through courts across the U.S. If those cases follow the same pattern, we haven’t seen the last of these headlines.

“Today, we are announcing an agreement with a bipartisan group of 52 attorneys general across US states, territories, and the District of Columbia, building on our longstanding efforts to empower parents and support teens.” reads the statement published by Meta.

“Over the years, we have consistently partnered with parents and experts — listening, learning, and building. That’s why we launched Teen Accounts in 2024, to bring automatic protections to teens, and more control for parents.”

The agreement aims to push YouTube, TikTok and other platforms to adopt similar protections for teenagers.

“While this is an important step, the fact is that teens move fluidly between dozens of apps a day. All platforms should empower parents and support teens by putting the same measures in place, because we know that when teens are restricted on one app, they simply move to another.” concludes Meta. “For meaningful progress to happen, we urge TikTok and YouTube to join us and state attorneys general in adopting this new standard, to ensure teens use social media in a healthy and responsible way.”

Follow me on Twitter: @securityaffairs and Facebook and Mastodon

Pierluigi Paganini

(SecurityAffairs – hacking, Facebook)

TikTok will pay $400 million to settle U.S. claims that it violated child privacy laws by collecting data from users under 13.

The U.S. Department of Justice announced that TikTok will pay $400 million to settle a 2024 lawsuit over children’s privacy.

“Today, the Department of Justice announced a $400 million settlement with TikTok, ByteDance, and affiliated entities (TikTok) resolving litigation concerning compliance with the Children’s Online Privacy Protection Act and its implementing regulations (COPPA).” reads the press release published by DoJ. “Under the settlement, TikTok will pay $300 million immediately and an additional $100 million upon entry of an order vacating a prior consent decree entered against TikTok’s predecessor, Musical.ly. The settlement represents one of the largest recoveries ever obtained in a COPPA case.”

TikTok will pay $300 million immediately and another $100 million after a court order removes an earlier consent decree involving Musical.ly. The 2024 case, brought by the DoJ and FTC, accused TikTok of knowingly allowing children under 13 to create accounts and illegally collecting data from children using Kids Mode.

Since the Justice Department filed its lawsuit against TikTok in 2024, the company has made major changes to its ownership, management, compliance, and privacy practices. It has also introduced stronger safeguards for younger users, improved age controls, and expanded parental oversight.

The DOJ said these measures have advanced the goals of its case and strengthened protections for millions of U.S. families. The settlement reflects a focus on practical results, securing a significant recovery while recognizing TikTok’s compliance improvements. The case was filed in California and handled by the DOJ’s Civil Division following a referral from the FTC.

“This settlement is a major victory for American children and parents,” said Associate Attorney General Stanley E. Woodward Jr. “The Department’s priority is ensuring that children are protected online and that companies entrusted with their personal information meet their legal obligations. This resolution secures a substantial recovery while reinforcing the protections that families expect and deserve.”

TikTok has faced regulatory scrutiny over children’s privacy before. In September 2023, Ireland’s Data Protection Commission fined the company €345 million for breaching the GDPR through its handling of children’s personal data.

Follow me on Twitter: @securityaffairs and Facebook and Mastodon

Pierluigi Paganini

(SecurityAffairs – hacking, privacy)

Trump authorizes vetted US cybersecurity firms to conduct government-approved cyber operations against transnational criminal networks.

President Trump signed a national security memorandum on August 13 establishing a formal program that allows vetted private US cybersecurity companies to conduct offensive cyber operations against transnational criminal organizations under government direction and oversight. The program, managed by the National Coordination Center, covers both intelligence collection, described as Cyber Surveillance Operations, and active disruption of criminal infrastructure, described as Cyber Effects Operations. It’s the formal implementation of what the White House’s Cyber Strategy for America promised in March: unleashing the private sector as an offensive cyber instrument.

“The American private sector is the most innovative and technologically advanced in the world, and its scale, speed, and capacity secure a critical offensive cyber advantage for the United States. Yet, American businesses’ innovative capabilities have historically been underutilized in efforts to identify and disrupt criminal networks operating in cyberspace. Thus, it is the policy of the United States to use all instruments of national power, including the innovative capabilities of the private sector, to combat cybercrime.” states the memorandum.

“By partnering with vetted United States companies subject to the direction and oversight of the Federal Government, we will enhance our ability to counter TCO threats and combat transnational cybercrime, fraud, and other predatory schemes against American citizens.”

The program targets what the memo defines as Cyber-Enabled Transnational Criminal Organizations, any foreign group conducting cyber-enabled crime against US interests, explicitly excluding entities that are institutional parts of foreign governments or wholly operated under foreign government direction. That carve-out matters: this program is aimed at criminal networks, not nation-state adversaries. The line between the two is often blurry in practice, but the memo establishes the presumption that a group is not government-directed unless clear intelligence says otherwise.

““Cyber Effects Operation” means activity conducted in or through the interdependent network of information technology infrastructure that includes the Internet, telecommunications networks, computers, information systems, industrial control systems, networks, and embedded processors and controllers that results in the manipulation, disruption, denial, degradation, or destruction of information systems, networks, physical or virtual infrastructure controlled by information systems, or information resident thereon.” continues the memorandum.

Program executive directors from the Department of Justice and the Department of Homeland Security must co-approve every operation in writing before any action is taken. Operations that could produce those Critical Outcomes require additional authorization beyond the program executive directors, an explicit acknowledgment that some cyber actions cross into territory governed by the laws of armed conflict.

Companies wanting to participate must clear rigorous vetting, demonstrate technical capability, submit to annual evaluations, and maintain a bond or escrow of at least $1 million that is forfeited if they violate their contract terms. The operational procedures are to be finalized within 60 days, and the Justice Department will review any operation that touches a US person or raises domestic constitutional questions. The legal question hovering over the whole program is whether the CFAA exemption for lawfully authorized government investigative activities extends to private companies acting under government contracts, a question no US court has yet answered. Jenner & Block lawyers noted the exemption likely applies when companies operate under direct government direction, but wouldn’t cover independent offensive operations without that oversight. That’s precisely why the memo makes government control explicit at every step: every operation needs written approval before action, every unintended contact with a US person or system must trigger an immediate stop and notification, and the Justice Department stays in the loop throughout.

Follow me on Twitter: @securityaffairs and Facebook and Mastodon

Pierluigi Paganini

(SecurityAffairs – hacking, Transnational Criminal Networks)

China opened a cybersecurity review of Palo Alto Networks, citing national security concerns but giving no details about the reasons behind the probe.

China’s Cyberspace Administration (CAC) announced that it’s launching a cybersecurity review of products Palo Alto Networks sells in the country. The announcement itself runs to a few sentences of formal Chinese, citing national security law and cybersecurity law as the basis for the review, and offers essentially nothing beyond that.

“To ensure the safe and stable operation of critical information infrastructure, prevent cybersecurity risks and vulnerabilities, and safeguard national security, in accordance with the National Security Law of the People’s Republic of China and the Cybersecurity Law of the People’s Republic of China, the Cybersecurity Review Office, following the Cybersecurity Review Measures, has conducted a cybersecurity review of Palo Alto Networks’ products sold in China.” the regulator says.

That’s the entire public justification. No specific vulnerability named, no incident referenced, no timeline for when findings might land. Palo Alto Networks told The Register it maintains high standards across its global operations and that, for now, there’s no impact on its ability to serve customers or deliver products in the region.

The situation resembles China’s 2023 security review of Micron, which was announced without warning. Weeks later, Beijing deemed Micron’s products a security risk for critical infrastructure, effectively restricting sales, but provided little explanation. Micron eventually withdrew its data center and server products from China, losing billions in annual revenue while local chipmakers gained new opportunities.

Micron’s story offers one genuinely reassuring data point for Palo Alto, if it’s any comfort: getting banned from a major market didn’t permanently damage the company. The AI boom drove memory prices high enough afterward that the China restriction barely shows up in Micron’s financials today. Getting frozen out of a market stings a lot less when the rest of the world is buying everything you can produce anyway.

China has already been pushing companies away from foreign cybersecurity products: in January, authorities reportedly told Chinese firms to stop using security software from a list of U.S. and Israeli vendors that included Palo Alto Networks, Fortinet, Check Point, CrowdStrike and others, encouraging the replacement of those products with domestic alternatives. Chinese vendors such as Huawei and H3C have increasingly positioned their own firewalls and security platforms as alternatives to foreign products, while other domestic companies are expanding across the cybersecurity market. H3C, for example, openly promotes its security products as replacements for overseas technologies.

That makes the Palo Alto review more than a simple technical investigation. It fits a broader strategy in which cybersecurity and national security are increasingly intertwined with China’s push for technological self-reliance.

For Palo Alto, the immediate financial impact is difficult to measure because the company does not separately report China revenue, but restrictions could still create an opening for domestic competitors while adding another layer of uncertainty for Western technology companies operating in the country.

There is also a wider geopolitical dimension. Beijing has repeatedly framed foreign technology as a potential national-security risk, while Washington and its allies have taken similar measures against Chinese and Russian vendors, including Huawei, ZTE and Kaspersky. The United States, for example, banned Kaspersky’s cybersecurity and antivirus products over national-security concerns, arguing that the software created risks because of its ties to Russia. The Palo Alto case therefore sits within a much larger cycle of reciprocal distrust, in which cybersecurity products are increasingly treated not only as commercial technologies but also as potential strategic assets.

Follow me on Twitter: @securityaffairs and Facebook and Mastodon

Pierluigi Paganini

(SecurityAffairs – hacking, China)

Meta was ordered to pay $567M after a judge ruled its platforms harmed children, bringing New Mexico penalties to $942M.

Meta ‘s child-safety legal bill just got another half-billion dollars heavier. A New Mexico state judge ruled that company’s platforms constitute a “public nuisance,” the BBC reports, ordering $567 million into a fund meant to address harm the company caused to children. Combined with an earlier $375 million penalty from the same case, Meta now owes New Mexico $942 million total.

“Judge Bryan Biedscheid said the social media giant is a “public nuisance” akin to air pollution and that it must put the money in a fund aimed at reducing future harms.Thursday’s ruling is in addition to $375m in fines Meta was already ordered to pay in the case, for a total of $942m.” BBC reports. “Judge Biedscheid compared Meta to a factory, with advertising and content as its product and “the psychological harm and sexual exploitation of children to be the pollution that must be abated”.”

Judge Bryan Biedscheid didn’t hold back on the framing. He compared Meta to a factory, with advertising and content as its output and the psychological harm and sexual exploitation of children as the pollution that output produces. It’s the kind of comparison a judge doesn’t reach for lightly, and according to CNN, it’s the first time any social media company has been legally labeled a public nuisance.

“The court found that “just as noxious pollution produced by the factory can harm the common public right to reasonably clean air, the harmful effects of Meta’s platforms on children do not stay contained by its platforms and, instead, migrate to the internet as a whole and, perhaps most concerning, to the real world and create a common, societal burden on and harm to the affected children and their families and schools, as well as hospitals and law enforcement.”” CNN reports.

The case traces back to a 2023 lawsuit from state attorneys general, and it unfolded in two phases. A March jury verdict already found Meta had repeatedly violated New Mexico’s Unfair Practices Act, largely because its recommendation algorithms steered young users toward harmful content and predatory contacts. This second phase, decided by the judge alone rather than a jury, existed specifically to answer one question: did that harm rise to the level of a public nuisance affecting the broader community.

According to CNBC’s reporting, Biedscheid’s written ruling didn’t pull punches on causation either.

“Expert testimony supports a causal link between social media and the youth mental health crisis in New Mexico,” the ruling states, closing off Meta’s usual argument that any correlation is just correlation.

Most of the money has a specific destination. $420 million goes toward direct treatment, funding clinical and behavioral health programs for young people already affected. The remainder covers prevention training for teachers and healthcare workers, plus broader awareness efforts, all running over roughly the next five years, according to PBS.

Cash isn’t the only thing Meta has to hand over. The judge ordered a list of concrete platform changes: no recommending accounts of users under 18 to adults, no adults messaging minors, a ban on sending or receiving nudity for underage accounts, and elimination of “like” counts for teen users. Push notifications get blocked overnight and during school hours on weekdays, and total monthly usage for minors gets capped at 90 hours across Instagram and Facebook combined, roughly three hours a day.

Meta’s response was predictable and brief. A company spokesperson said Meta disagrees with the ruling and will appeal, adding that the company has worked hard to keep people safe and remains confident in its record protecting teens online.

“We disagree with the ruling and will appeal.” a company spokesman told BBC. “We work hard to keep people safe on our platforms and have been transparent about the challenges of identifying and removing bad actors and harmful content,” he added.

“We remain confident in our record of protecting teens online and will continue to defend ourselves against claims that misrepresent the facts.”

That’s the same basic line the company used after the March verdict, and it’s likely to stay the company line through however many appeals this takes.

New Mexico is far from the only front in this fight. Nearly three dozen state attorneys general are pursuing a separate case against Meta over child privacy violations, with another major trial starting next week in California, and Meta already lost a Los Angeles case earlier this year that found it could be held liable for building deliberately addictive platforms. Add in the EU’s ongoing preliminary findings against Meta over underage users on Instagram and Facebook, and the pattern stops looking like isolated lawsuits and starts looking like a coordinated reckoning across multiple jurisdictions at once.

Former Twitter executive Bruce Daisley put the number in context on BBC Radio 4, calling it “a drop in the ocean” against Meta’s finances; the company posted $61 billion in quarterly revenue this year, up 28% from the year before. The fine is real money by any normal measure. Whether it’s real money by Meta’s measure is a different question entirely, and it’s the one regulators worldwide are now racing to answer with policy rather than just penalties.

Follow me on Twitter: @securityaffairs and Facebook and Mastodon

Pierluigi Paganini

(SecurityAffairs – hacking, Meta)

Qilin ransomware exploits the PAN-OS GlobalProtect flaw CVE-2026-0257 to gain unauthorized VPN access to unpatched networks.

Arctic Wolf researchers warn that the Qilin ransomware gang is exploiting the critical PAN-OS GlobalProtect vulnerability CVE-2026-0257 to compromise corporate networks.

CVE-2026-0257 is a PAN-OS authentication bypass vulnerability affecting GlobalProtect portals and gateways.

Palo Alto Networks addressed the vulnerability on May 13. Two weeks later, cybersecurity firm Rapid7 confirmed active exploitation across multiple customer environments. In early June, the U.S. Cybersecurity and Infrastructure Security Agency (CISA) added the flaw CVE-2026-0257 to its Known Exploited Vulnerabilities (KEV) catalog.

The vulnerability affects the GlobalProtect portal and gateway components of Palo Alto Networks PAN-OS, allowing attackers to bypass authentication and establish unauthorized VPN connections. The vulnerabilities do not affect Panorama or Cloud NGFW deployments.

“Authentication bypass vulnerabilities in the GlobalProtect portal and gateway of Palo Alto Networks PAN-OS® software allows the attacker to bypass security restrictions and establish an unauthorized VPN connection.” reads the advisory.

Arctic Wolf warns that the Qilin ransomware gang is exploiting the critical PAN-OS GlobalProtect vulnerability CVE-2026-0257 to compromise corporate networks. The flaw allows attackers to bypass authentication and establish unauthorized VPN sessions on unpatched devices. Palo Alto Networks released patches on May 13 and confirmed exploitation attempts against systems that had not applied updates or mitigations.

Arctic Wolf Labs has observed several attacks in which threat actors exploited CVE-2026-0257 to gain initial access and deploy Qilin ransomware across entire Windows domains. Investigators found evidence that multiple Qilin affiliates are actively abusing the flaw to compromise organizations, making unpatched PAN-OS GlobalProtect devices a high-priority target for ransomware operations.

“Arctic Wolf investigated multiple distinct intrusions during June 2026 that resulted in Qilin ransomware deployment, all originating from exploitation of CVE-2026-0257 against Palo Alto Networks firewall appliances.” reads the report published by Arctic Wolf. “Post-exploitation tradecraft varied across intrusions, from rapid encryption-only operations to full double-extortion, possibly suggesting multiple affiliates operating under the Qilin ransomware-as-a-service (RaaS) umbrella.”

Arctic Wolf found that attacks exploiting CVE-2026-0257 followed a common initial pattern but diverged after compromise. Threat actors consistently used the same entry point, ransomware staging paths, PsExec execution, and registry persistence. However, some attacks quickly encrypted entire environments without stealing data, while others involved extensive reconnaissance, deployment of remote-access tools such as AnyDesk, Ngrok, and LogMeIn, large-scale credential theft, and data exfiltration to cloud services before ransomware execution, reflecting the varied tactics of Qilin RaaS affiliates.

After exploiting CVE-2026-0257, the attackers established VPN sessions from Kali Linux systems, then quickly secured persistent access using registry Run keys, scheduled tasks, and remote administration tools such as AnyDesk, Ngrok, LogMeIn, and MeshAgent. They harvested credentials by dumping LSASS memory and extracting the Active Directory database (NTDS), enabling lateral movement with PsExec, RDP, and compromised administrator accounts.

The operators scanned networks with SoftPerfect Network Scanner and NetExec, cleared Windows event logs, and in some cases disabled Microsoft Defender before deploying ransomware. Several intrusions also involved data theft using Rclone, ProtonDrive, FileZilla, and MEGA cloud storage, while others focused solely on rapid encryption.

The ransomware payload, typically named win.exe, was staged in C:\PerfLogs, executed with password-protected parameters, and encrypted files using unique extensions assigned to each campaign.

“The variability in post-exploitation tradecraft, from encryption-only operations to full double-extortion, shows that perimeter compromise is the critical point for defenders. After exploitation succeeds, the impact depends on the affiliate’s goals and timeline, but domain compromise and ransomware deployment are consistent.” concludes the report. “Arctic Wolf Labs assesses with moderate confidence that intrusions leveraging CVE-2026-0257 and leading to Qilin ransomware deployment are likely ongoing. This assessment is based on the extensive scanning activity observed and the RaaS model’s tendency to distribute successful exploits among multiple affiliates.”

Qilin ransomware operation has been active since 2022, it has become one of the most active RaaS groups in 2025, claiming over 40 victims monthly and peaking at 100 in June.

The group enables affiliates to deploy customized ransomware payloads against targeted organizations. Qilin uses double-extortion tactics, encrypting data while threatening to leak it via Tor-based portals. The group has targeted multiple sectors worldwide, including healthcare, manufacturing, and finance, leveraging phishing and known vulnerabilities.

In October 2025, Resecurity’s researchers detailed how the Qilin RaaS group relies on global bulletproof hosting networks to support its extortion operations.

In early October, DragonForce, LockBit, and Qilin formed a ransomware alliance to boost attack effectiveness, marking a major shift in the cyber threat landscape. Ransomware groups DragonForce, LockBit, and Qilin formed a strategic alliance to enhance their attack capabilities, signaling an evolving cyber threat landscape. The alliance aims at sharing tools and infrastructure to enhance attack effectiveness. 

At the end of March, Qilin Ransomware group allegedly breached the chemical manufacturing giant Dow Inc. 

Follow me on Twitter: @securityaffairs and Facebook and Mastodon

Pierluigi Paganini

(SecurityAffairs – hacking, newsletter)

Bad Epoll (CVE-2026-46242) lets local attackers gain root on Linux and Android. The flaw was missed by AI but found by a security researcher.

A newly disclosed Linux kernel vulnerability, named Bad Epoll (CVE-2026-46242), allows a local attacker with no special privileges to gain full root access on affected Linux systems and Android devices. Security updates are already available, and users are urged to install them as soon as possible.

The flaw affects the Linux kernel’s epoll subsystem, a core feature used by servers, browsers, and countless applications to efficiently manage multiple network connections and file events. Because epoll is fundamental to Linux, there is no practical workaround other than patching vulnerable systems.

Bad Epoll is a classic use-after-free vulnerability, which occurs when a program continues to use a piece of memory after it has already been released (“freed”).

Two kernel threads attempt to release the same internal object simultaneously. One frees the memory while the other continues using it, creating a brief opportunity to corrupt kernel memory and escalate privileges to root.

“Two of epoll’s close paths run at the same time and collide. One frees an object while the other is still writing into it, and that is the use-after-free (UAF).” continues the advisory. “The race window, and how the exploit drives it. The exploit uses four epoll objects grouped into two pairs. One pair triggers the race, while the other becomes the victim. From there, the exploit turns the 8-byte UAF write into a UAF on a file object, and uses a cross-cache attack to fully control the file’s contents. Turning the bug into an arbitrary kernel memory read through /proc/self/fdinfo. With that control, the exploit gains an arbitrary read of kernel memory through /proc/self/fdinfo. Finally, it hijacks control flow and executes a ROP chain to gain a root shell.”

Although exploiting the flaw requires hitting a timing window only six CPU instructions wide, researcher Jaeyoung Chung developed a reliable proof-of-concept that reportedly succeeds in about 99% of attempts on tested systems. According to the researcher, the exploit can even be launched from Chrome’s renderer sandbox, making it particularly dangerous, and could also impact Android devices.

“Bad Epoll (CVE-2026-46242) is a race-condition use-after-free in the Linux kernel’s epoll subsystem. This bug lets an unprivileged process become root, not only on Linux desktops and servers but also on Android devices.” reads an advisory published by Chung.

One of the most interesting aspects of the vulnerability is its connection to AI-assisted vulnerability research. Bad Epoll originates from the same section of kernel code where Anthropic’s Mythos model previously identified another privilege escalation flaw, tracked as CVE-2026-43074. The AI detected the first bug, but missed this closely related vulnerability, which was later discovered manually.

“A single commit in 2023 introduced two separate race conditions into the epoll code, only about 2,500 lines in all. Both turned out to be critical bugs that can lead to privilege escalation.

The first was found by Anthropic’s Mythos and reported as CVE-2026-43074. That result is impressive on its own, because kernel race bugs are known to be hard to find. It showed a frontier AI model’s ability to find race bugs. An independent researcher later submitted a 1-day exploit for it to kernelCTF.” continunes the advisory. “The other race is Bad Epoll, which Mythos missed.”

Chung believes the miss is understandable. The race condition is extremely difficult to reason about because the vulnerable execution path exists for only a tiny fraction of a second. In addition, once the first flaw was patched, Bad Epoll no longer generated obvious warnings through KASAN, Linux’s memory error detection system, making it even harder to spot.

The good news is that there is currently no evidence that Bad Epoll has been exploited in the wild. The only public exploit is the proof-of-concept released through Google’s kernelCTF program. An Android exploit is reportedly still under development.

The flaw affects Linux kernels based on version 6.4 and later, unless they already include the upstream fix. Older long-term support kernels based on Linux 6.1, including some Android devices such as the Pixel 8, are not vulnerable because the problematic code was introduced after those versions branched.

Bad Epoll joins a growing list of high-profile Linux privilege escalation vulnerabilities recently disclosed, including Copy Fail, Dirty Frag, Fragnesia, and DirtyClone. While many of these newer vulnerabilities are deterministic and relatively easy to exploit, Bad Epoll belongs to the older class of race-condition bugs, which are significantly harder to discover, exploit, and patch.

The case also highlights both the promise and the current limitations of AI in vulnerability research. Models such as Mythos have already demonstrated they can identify complex kernel flaws and even uncover long-standing vulnerabilities in projects like FreeBSD.

At the same time, Bad Epoll shows that highly subtle race conditions can still escape even state-of-the-art AI systems. For now, human expertise remains essential, particularly when vulnerabilities depend on tiny timing windows and complex concurrent execution paths.

Follow me on Twitter: @securityaffairs and Facebook and Mastodon

Pierluigi Paganini

(SecurityAffairs – hacking, Bad Epoll)