Anthropic said its month-old Project Glasswing initiative has uncovered more than 10,000 high- or critical-severity software vulnerabilities across systemically important code, a finding the company says has shifted the central problem in cybersecurity from discovering flaws to verifying and patching them.

The findings, drawn from partner reports and independent evaluations, mark one of the first large-scale accountings of what a frontier AI model can do when pointed at widely used code, and of the bottlenecks that emerge once it does.

Several partners reported that their rates of bug discovery had increased more than tenfold. Cloudflare identified 2,000 bugs across its critical-path systems, including 400 rated high or critical, with a false-positive rate the company said it considered better than that of human testers. At one unnamed partner bank, the model was credited with helping detect and prevent a fraudulent $1.5 million wire transfer initiated after a customer’s email account was compromised and followed up with spoofed phone calls.

External evaluations cited in the update tracked with the results Anthropic released. The United Kingdom’s AI Security Institute found that Mythos Preview was the first model to solve both of its cyber ranges — simulations of multistep cyberattacks — from end to end. Mozilla said it found and fixed 271 vulnerabilities in Firefox 150 while testing the model, more than 10 times the number found in Firefox 148 using an earlier Anthropic model. AI-powered security platform XBOW called the model a significant step up over existing systems on its web exploit benchmark.

Anthropic also used Mythos to scan more than 1,000 open-source projects. The model has flagged 23,019 potential vulnerabilities, 6,202 of them estimated as high or critical. Of 1,752 high- or critical-rated findings reviewed by six independent security research firms or by Anthropic itself, over 90% were confirmed as valid, and over 62% were confirmed to be high or critical.

The company did note that while it’s good at finding vulnerabilities, there is still a gap in having people fix every issue. 

“The bottleneck in fixing bugs like these is the human capacity to triage, report, and design and deploy patches for them,” the report states. 

Open-source maintainers have also been contending with a wave of low-quality, AI-generated bug reports, and Anthropic said it tries to reproduce and assess each issue before reporting it. At maintainers’ request, it has sometimes disclosed bugs without further vetting, reporting 1,129 such cases, of which the model estimated 175 to be high or critical.

Anthropic said it has not released Mythos-class models publicly because no company, including itself, has developed safeguards to prevent serious misuse. In the interim, it has released Claude Security in public beta for enterprise customers, which it said has been used to patch more than 2,100 vulnerabilities in three weeks using the publicly available Claude Opus 4.7, and has begun a Cyber Verification Program for security professionals.

The company said it plans to expand Project Glasswing with additional partners, including U.S. and allied governments, before any broader release of the underlying model.

“Glasswing helps the most systemically important cyber defenders gain an asymmetric advantage. However, there is an urgent need for as many organizations as possible to shore up their cyber defenses,” the report states. “We hope that our generally available models, and the new tools, resources, and research we’re providing to accompany them, will support those organizations to improve their cybersecurity posture.”

The post Anthropic: Mythos finds more than 10,000 software flaws in first month appeared first on CyberScoop.

Authorities seized infrastructure powering four botnets that hijacked a combined three million devices and launched more than 300,000 DDoS attacks collectively, the Justice Department said Thursday.

The botnets — Aisuru, Kimwolf, JackSkid and Mossad — enabled operators to sell access to the infected devices for various cybercrimes. The aftermath spanned thousands of attacks, including some demanding extortion payments from victims, officials said.

The globally coordinated operation, aided by law enforcement actions targeting the botnets’ operators in Canada and Germany, disrupted the command-and-control infrastructure for all four botnets. Two of the botnets set records before the takedown, attracting widespread attention from security researchers and vendors.

The Kimwolf botnet, an Android variant of Aisuru, spread like wildfire after its operators figured out how to abuse residential-proxy networks for local control, according to Sythient. It eventually took over more than 2 million Android TV devices by January. In September, just as Kimwolf was forming, Cloudflare clocked the Aisuru botnet hitting a record-breaking 29.7 terabits-per-second DDoS attack that lasted 69 seconds.

Officials ultimately attributed roughly 200,000 DDoS attacks to Aisuru, 90,000 to JackSkid, 25,000 to Kimwolf and about 1,000 DDoS attack commands to the Mossad botnet. Yet, DDoS attacks from financially-motivated attackers are typically a distraction or misdirection.

“Oftentimes a DDoS attack is just advertising for the size of an operator’s botnet,” Zach Edwards, staff threat researcher at Infoblox, told CyberScoop. Botnet operators cash out by renting these controlled devices to cybercriminals for account abuse, password reset attacks, ad fraud schemes and residential proxy nodes, he added.

Devices infected by the four botnets include digital video recorders, web cameras, Wi-Fi routers and TV boxes. Hundreds of thousands of these devices are located in the United States, federal prosecutors said. 

Authorities did not name the people involved or formally announce any arrests. Yet, they describe the operation in nearly conclusive terms, claiming the action disrupted the botnets’ communications infrastructure — domains, virtual servers and other systems — to prevent further infection and limit or eliminate the botnets’ ability to launch future attacks.

“Cybercriminals infiltrate infrastructure beyond physical borders and Defense Criminal Investigative Service participates in international operations to help safeguard the Department’s global footprint,” Kenneth DeChellis, special agent in charge at the Defense Department’s DCIS cyber field office, said in a statement. Some of the DDoS attacks attributed to these botnets reached IP’s owned by the Department of Defense Information Network.

Botnets often compete for devices to infect and opportunities to scale. As Kimwolf spread and hit those objectives, it captured sweeping interest from researchers, authorities and vendors in a position to help stop it. 

Kimwolf was the largest DDoS botnet ever detected, according to Tom Scholl, vice president at Amazon Web Services, which assisted the operation. “The scale of this botnet is staggering,” he said in a LinkedIn post. 

“Kimwolf represented a fundamental shift in how botnets operate and scale,” Scholl added. “Unlike traditional botnets that scan the open internet for vulnerable devices, Kimwolf exploited a novel attack vector: residential proxy networks.”

Under this mechanism, any organization with vulnerable devices connected to the internet could unwittingly have those devices turned into a node for a botnet or a foothold for a targeted attack.

“This isn’t just some problem that your cousin has because he bought some cheap TV box that promised him free TV channels,” Edwards said. Infoblox previously said nearly 25% of customers had at least one endpoint device in a residential proxy service targeted by Kimwolf.

While it’s intellectually interesting whenever a botnet scales to extraordinary size, it’s also a “sad reminder that oftentimes security takes a back seat to convenience and cost,” Edwards said. 

“The botnets are growing because more and more people are buying weird internet-connected stuff,” he added. “Nothing in this world is free.”

The takedowns mark a continuation of a consistent, ongoing crackdown targeting large-scale botnets, cybercrime marketplaces, malware, infostealers and other cybercrime tools. Some of the malicious networks hampered or rendered nonoperational by disruptions and arrests during the past year include: DanaBot, Rapper Bot, Lumma Stealer, AVCheck and SocksEscort.

More than 20 companies and organizations assisted with the coordinated disruption, including law enforcement from the Netherlands and Europol. Efforts to stop botnets will continue as these malicious networks proliferate in new places and new ways. 

“We’re living in a device-compromise–DDOS-botnet-merry-go-round and while many of us wish something could slow it down, the challenges continue to grow,” Edwards said. “This is still a bad day for serious threat actors, and any day like that is something we should all celebrate.”

The post Justice Department disrupts botnet networks that hijacked 3 million devices appeared first on CyberScoop.