A Supply Chain Attack Put Malicious Anthropic Claude Extensions on Thousands of Devices
Thanks to a lucky break, flawed code in a tainted Claude extension prevented the attacker from fully worming their way into some 5,000 developer machines.
Thanks to a lucky break, flawed code in a tainted Claude extension prevented the attacker from fully worming their way into some 5,000 developer machines.
A deadly attack on a Christian funeral in Jordan and subsequent Jordanian airstrikes in Syria have sparked a diplomatic dispute between Amman and Damascus. Bill Roggio is joined by FDD Senior Research Analyst Ahmad Sharawi to discuss the Islamic State's possible involvement, the deteriorating security situation in southern Syria, and the implications for US counterterrorism policy in the region.
Recognized as a top threat to AI agents, prompt injection allows instructions from untrusted data (email, documents, websites) to manipulate the agent's operation. The existing best-defended LLM, Meta-SecAlign, is trained with the DPO loss but still suffers from attack success rates (ASRs) approaching 100% under adaptive prompt injections. One reason is that DPO anchors the model around the undefended reference model, which is insecure. We show an idealized reference…
RAID-AI aims to help organizations assess their security preparation for ransomware attacks targeting AI systems
Over 22.6 million exposed instances are potentially affected by the critical PHP flaws that have been targeted for over a week.
Germany’s role as a central pillar of the alliance confronting Russia is under attack from the extremes. It cannot be ignored by the main centrist parties.
A security firm patched a decade-old app to prevent the takeover of millions of machines, highlighting that abandoned domains remain a boon to attackers.
As organizations eliminate passwords, attackers will adapt. Instead of attacking the authenticator directly, they will look for opportunities around it.
A 15-year-old student in Gainesville, Florida, allegedly sought guidance from Islamic State members and supporters on building drones and explosives to attack people at an unidentified indoor location.
The Interior Ministry confirmed that one person has been arrested, as Russian sources cited a tourist from Belarus alleging that a man physically attacked her at Tbilisi railway station because she was speaking Russian. The alleged assailant reportedly claimed that he hit her after she rejected his attempt to introduce himself. The incident, which reportedly …
Threat actors are attempting to exploit a high-severity unauthenticated race condition leading to remote code execution.

Each day the ceasefire in Oleshky holds means hope for civilians in the city, the UN’s top humanitarian official said, stressing that those wishing to leave must be able to do so safely and voluntarily.
Israeli forces continue to push Palestinians into ever-smaller areas of the Gaza Strip while destroying civilian infrastructure and attacking displacement sites, the UN human rights office, OHCHR, warned on Friday amid concerns of ethnic cleansing.
CISA has added two new vulnerabilities to its Known Exploited Vulnerabilities (KEV) Catalog, based on evidence of active exploitation. CVE-2026-111357 Ivanti Connect Secure Missing Authentication for Critical Function Vulnerability CVE-2026-118527 Ivanti Connect Secure Generation of Predictable Numbers or Identifiers Vulnerability These types of vulnerabilities are a frequent attack vector for malicious cyber actors and pose significant risks to the federal enterprise…
Leading cryptocurrency analysts have attributed the cyber-attack to a North Korean elite hacking unit
Malicious code linked to a website monitoring tool was designed to hijack logged-in Polymarket sessions and siphon user funds.
Tracked as CVE-2026-97229, the bug allows unauthenticated attackers to execute arbitrary code on the server.
The "FrostyNeighbor" advanced persistent threat (APT) launched a destructive attack in Ukraine and is raising concerns about deeper alignment with Russian cyber objectives.
Multimodal indirect prompt injection embeds malicious instructions in external media that also contains information needed for the user's task. Existing defenses can block the attack yet impair the model's use of that information, leaving a gap between attack defense and task preservation. We propose ERASE, an on-policy self-distillation method that constructs its reference by erasing the injected region in place. Modality-appropriate blank content replaces the injection…
North Korea’s buildup along the DMZ is designed to entrench sovereignty, not launch an attack.
The rapid adoption of digital technologies, online learning platforms, social media, cloud services, digital payments, and artificial intelligence has transformed higher education while increasing exposure to cyber threats. This review-based paper examines cyber-attacks among higher education students through an integrated conceptual Structural Equation Model (SEM) organized into three interconnected dimensions: WHAT, WHY, and HOW. The WHAT dimension explains the nature…
Lamis was 8. Her brothers, Osama and Yaseen, were 9 and 11. They were killed alongside their uncle and grandmother when Jordanian jets struck their village in southern Syria, in what Amman called a precision attack against the Islamic State group. The children’s father had been disappeared by Assad.
Ambulances were diverted, surgeries cancelled, and medical students sent home from classes after a ransomware group attacked the University of New Mexico's Health Sciences Center (HSC) this week.
Despite growing capabilities, LLMs remain susceptible to prompt injections: malicious inputs that can manipulate or hijack models. In particular, it has recently been observed that LLMs can exhibit role confusion (Ye et al., 2026), where models attribute the source of text based on its style instead of its labeled role. Malicious text in tool outputs (for instance, a malicious webpage accessed by an agent) can leverage role confusion to impersonate the user, an attack we…
Cisco open sourced Project Quarry, a rule language for detecting AI agent attacks that only emerge across sequences of benign actions.
Experts examine four major international legal questions raised by Russian and Ukrainian attacks on data centers.
Amman, October 29, 2026 — The Committee to Protect Journalists condemns the detention of Russian-Israeli journalist Andrey Khrzhanovskiy, known professionally as Andrey X, and U.S. journalist Jasper Nathaniel by Israeli forces near Umm Al-Khair, in the occupied West Bank. The arrests followed a violent assault by Israeli settlers against the two reporters and fellow freelance...

The UN’s top humanitarian official in Ukraine on Thursday condemned a Russian drone attack on the city of Dnipro that killed at least 13 civilians and injured 10 others at its main railway station.

At least five children were killed and two others injured in a drone attack on a market in Nyala, in Sudan’s South Darfur state, on Monday, the UN Children’s Fund (UNICEF) said on Thursday.