Vulnerability | Breaking Cybersecurity News | The Hacker News

Anthropic on Friday said it's cutting off live internet access for all its internal evaluations following the discovery of new incidents in which its artificial intelligence (AI) models exhibited misaligned behavior and targeted real websites. The AI company said it identified four broad categories of unintended model actions during evaluations and internal use of Claude - Claude Mythos Preview exploiting SQL or command injection flaws in unspecified third-party software to run commands on a university server, either because its own tools were intentionally limited or because an outside service it needed was unavailable, causing it to use other tools hosted on a third party's site to complete the task. Claude Haiku 4.5 and a non-frontier research model submitting a sensitive form on a real website when it was not authorized to do so. This occurred in scenarios where instructions were ambiguous or due to environment misconfigurations that prevented the agent from w...

Cybersecurity researchers have disclosed details of a previously unseen variant of the DarkSword iOS exploit kit called P7 DarkSword . "Compared with the variants we usually observe, P7 reduces its on-device footprint, adds on-device keychain and crypto-wallet theft, and adds two way C2 communication with the attacker's infrastructure," iVerify said in a new report published Thursday. The name "P7" is a nod to the threat actor's use of the "p7_" variable prefix in changes made to the original DarkSword code. DarkSword was first publicly documented earlier this March by Google Threat Intelligence Group (GTIG), iVerify, and Lookout, detailing its ability to target iPhones running iOS versions between iOS 18.4 and 18.7. The kit was detected in the wild in November 2025. The toolkit is engineered to chain multiple iOS vulnerabilities to escape the browser sandbox, escalate to kernel privileges, and inject the main payload into SpringBoard, t...

Four more U.S. states sued router maker TP-Link Systems on October 6, bringing the total to five, with   Texas filing a suit in February . Florida, Iowa, Montana and Nebraska allege the California company misled buyers about how secure its routers are and how separate it is from China. TP-Link  denies the claims  and says it will fight them in court. TP-Link Systems is based in Irvine, California. Until a 2024 restructuring, it was affiliated with TP-Link Technologies, a Chinese company that the suits do not name as a defendant. The complaints from  Florida ,  Montana  , and  Nebraska  do not allege that the Chinese government has obtained customers' data through TP-Link. They describe that as a risk under Chinese law. Separately, they say state-backed hackers have exploited flaws in TP-Link routers. Iowa's announcement is worded more strongly in places. Attorney General Brenna Bird's office said TP-Link firmware gives the Chinese governm...

Security researchers have  published a full working exploit  for a pre-authentication remote code execution flaw in AnyDesk Linux that gives attackers root access before anyone approves the connection. AnyDesk patched the flaw in version 8.0.3 in June, but its  changelog  described the fix only as "fixed a bug that could lead to a crash," with no CVE assigned and no security advisory. The exploit, called AnyPwn, targets a heap buffer overflow in AnyDesk's session protocol, a remote desktop tool. The code was released on GitHub on October 8. Administrators should update AnyDesk Linux to at least version 8.0.3. The latest release is 8.1.0. What the Exploit Demonstrates The published exploit works only over direct TCP connections on port 7070. The exploit is probabilistic: the heap layout must place a target object adjacent to the overflowed buffer; otherwise, the service crashes instead of executing the attacker's command. The offsets in the published code...

Anthropic on Thursday unveiled OSS Scanner as an opt-in vulnerability scanner to help secure the open-source ecosystem using artificial intelligence (AI). "It's an opt-in service informed by our experience using Claude to find vulnerabilities during Project Glasswing," Anthropic said . "Projects that join will receive thorough, periodic security scans by our strongest models at no cost." Anthropic also noted that the outputs of the scanner will be fully model-generated and do not require human review or triage, thereby facilitating faster and more frequent scanning. These reports are expected to be generated by its strongest models, including Claude Mythos. The company pointed out that it expects to use a set of criteria similar to Google's OSS-Fuzz to pick projects, while emphasizing that the process may evolve over time. Project maintainers are advised to provide a short description  explaining the importance of their project in cases where "...