Latest cybersecurity reporting from selected sources.

Claroty Warns of Downtime, 'Devastating' Impact of Vulnerabilities in OT SystemsVulnerabilities in backup power devices and heating and cooling control systems widely used in data centers could enable remote cyberattacks and result in costly downtime and "devastating" operational impact, according to new research from OT security firm Claroty.

Study Finds Standard Safety Tests Miss Most Agentic AI ThreatsResearchers from Stanford, MIT, Carnegie Mellon and others found that most production AI agents are vulnerable to attacks that unfold across multi-step actions. The study warns that memory, tool access and agent coordination create failure modes traditional chatbot safety testing cannot see.

AI Dependency Attack Reportedly Exposes Data and Source CodeA LiteLLM supply-chain compromise enabled attackers to harvest credentials and access internal environments at scale at Mercor. The firm was the first to confirm a LiteLLM breach, and researchers are warning about growing AI system exposure and limited visibility.

Also, CIRO Phishing Breach, Ingram Micro Ransomware and CVE SurgeThis week, DOGE posted sensitive data on an outside server. A phishing attack affected 750,000 Canadians. A hacktivism warning from the U.K. NCSC. An Ingram Micro breach. CVEs surged in 2025. SK Telecom challenged a fine. Researchers disclosed Chainlit flaws. North Korean hackers abused VS Code.

Foundation for Defense of Democracies Warns Against Aligning With New Cyber TreatyThe United Nations' cybercrime treaty, shaped by Russian and Chinese influence, could legitimize global digital repression by enabling prosecutions of journalists, activists and researchers under vague terms - despite U.S. opposition and mounting civil society alarm, analysts warned Thursday.

Today's Hapless Hackers Are Tomorrow's Threat, Warns ForescoutA pro-Russian hacktivist group boasted on Telegram that it hacked a Western water treatment plant - but actually succeeded in attacking a honeypot left by security researchers at Forescout, the firm said. TwoNet appears to have ceased operations on Sept. 30.

Today's Hapless Hackers Are Tomorrow's Threat, Warns ForescoutA pro-Russian hacktivist group boasted on Telegram that it hacked a Western water treatment plant - but actually succeeded in attacking a honeypot left by security researchers at Forescout, the firm said. TwoNet appears to have ceased operations on Sept. 30.

'More Research' Needed, Says GoogleGoogle DeepMind expanded its risk framework to cover scenarios where artificial intelligence models might manipulate people or resist shutdown, marking the company's most explicit warning yet about potential misalignment. DeepMind frames many of these scenarios as a malfunction.

Robust Security Reviews of AI-Generated Code Required, Warns StudyGenies might grant magic wishes, but they inevitably come with consequences. Enter artificial intelligence coding assistants, which many developers say help them write code more quickly. New research shows that this productivity comes with a security cost, and must be mitigated.

Semperis Warns of Flaw in Windows Server 2025 Delegated Managed Service AccountsA critical cryptographic flaw in Windows Server 2025's delegated Managed Service Accounts, or dMSAs, allows attackers to generate passwords for every managed service account across an Active Directory forest and create a backdoor, Semperis researchers found.

Experts Aim to Probe How AI Models Reason, and Why It MattersAI researchers from OpenAI, Google DeepMind and Anthropic and others have urged deeper study into chain-of-thought monitoring, a technique to track how reasoning models arrive at answers. Their joint paper warns that transparency may erode if not prioritized.

Always Secure MCP Servers Connecting LLMs to External Systems, Experts WarnWarning: Popular technology designed to make it easy for artificial intelligence tools to connect with external applications and data sources can be turned to malicious use. Researchers discovered two separate vulnerabilities tied to tools in the ecosystem around model context protocol, or MCP.

Research Shows How Large Language Models Fake Conceptual MasteryMIT, Harvard and University of Chicago researchers say models suffer from "potemkin understanding," referring to an illusion where models ace conceptual tests but fail real-world application. Their paper warns this undermines benchmarks and points to gaps in genuine AI comprehension.

Google’s Mandiant Warns About Remote Attacks Disrupting Grid StabilityVulnerabilities in networked devices programmed to instantaneously trip power grid substation circuit breakers could be the means hackers use to cause the next blackout, warn researchers. There are "systemic patterns across substations, utilities and industrial sites worldwide," Mandiant warned.

'Heightened Threat Environment' Faces Critical Infrastructure, US Government WarnsMany types of commonly used types of industrial control systems continue to be deployed in a manner that leaves them publicly exposed to the internet, often by U.S.-based critical infrastructure operators, in what amounts to a preventable security risk, researchers warn.

Research Shows Next-Generation 9-1-1 Ecosystems Lack Critical Cyber ProtectionsA report from telecom firm Intrado warns that cybersecurity safeguards are lagging behind the rapid deployment of next-generation 911 systems, exposing the emergency ecosystem to attacks ranging from VoIP floods to ransomware amid growing reliance on cloud-based and IP-connected technologies.

Defunct Ransomware Group's Diaspora Includes Hackers With Focus on Microsoft TeamsBased on intelligence gleaned from the leak of Black Basta ransomware messages, researchers are warning organizations to beware phishing attacks launched via Microsoft partner domains and via Teams, as well as the targeting of personal Google accounts accessed via corporate devices.

Akamai Researchers Flag 'BadSucessor' in Windows Server 2025An unpatched flaw in Windows Server 2025 that is "trivial" to exploit and present in the default configuration is full domain compromised, warns new research from Akamai. The flaw is present in a new account type known as delegated managed service accounts, or dMSA.

Researchers Say Chinese Mobile Route Firms Dominate Global Interconnect IndustryA report warns U.S. allies and countries across the globe are using Chinese-owned and controlled mobile routing firms in a move that could risk national security interests and potentially expose billions of users to passive and active surveillance from Beijing.

Cybersecurity Firm Warns of 'Hidden Features' in the Widely Used ESP32 ChipResearchers have discovered 29 undocumented commands in the Bluetooth implementation of the ESP32 microcontroller - one of the world's most widely used chips in internet of things devices and industrial IoT devices - that attackers could use to read and modify device memory.