artificial intelligence | Breaking Cybersecurity News | The Hacker News

A package gets installed. A login prompt opens. A box sits exposed to the internet. Nothing looks unusual yet. That’s roughly the mood this week. Trusted tools turn hostile, old weak spots get fresh attention, AI makes exploit work cheaper, and researchers keep finding attacks that sound harder than they actually are. Plenty to clean up. Here’s the short version. ⚡ Threat of the Week U.S. Warns of AI-Powered Attacks on Siemens PLCs — Threat actors are using AI to write exploit scripts targeting internet-exposed Siemens S7 Series programmable logic controllers (PLCs) used across water, energy, manufacturing, and other critical infrastructure sectors, according to the U.S. government. The agencies warned: "This is not a theoretical risk—it is an active threat." The exploitation of poorly secured PLCs could result in disruption of critical industrial processes, safety incidents, downtime or equipment damage, compromise of sensitive data, and compliance violations, not...

Big security risks come in small packages. While enterprise security teams focus on policing the proliferation of employees using ChatGPT and Claude for quick drafting tasks, a more urgent threat is posed by a handful of AI super-adopters who are quietly hardcoding unvetted tools into critical business operations. According to new research published by Akamai, the top 5% of enterprise power users interact with AI models at 12 times the rate of the bottom 50% of the workforce. The findings, published in Akamai’s State of the Internet: Enterprise AI Usage Risk Report 2026 , are based on real-world usage and telemetry data as well as research and threat analysis.  These AI super-adopters are creating a disproportionate security risk by expanding the use of shadow AI, increasing opportunities for data leakage, and introducing autonomous AI agents that operate within the enterprise but outside its established guardrails. “Small groups of AI power users are casting outsized shado...

Cybersecurity researchers have disclosed details of a Chinese-speaking cybercrime group dubbed UAT-10147 that's targeting Windows and Linux web servers globally across the education, media, technology, and gaming sectors. The vast majority of the targets are located in Brazil, Bolivia, China, Canada, and Vietnam. Details of the threat activity came to light following the discovery of an open directory hosted at "139.180.197[.]150," which was observed communicating with one of the compromised machines. "The actor leveraged publicly disclosed vulnerabilities to gain initial access at scale," Cisco Talos said in a two-part report published last week. The actor employed a mixture of open-source offensive frameworks, including Metasploit , ysoserial , PentestGPT , DeepAudit, and multiple privilege escalation exploits to automate intrusion operations and establish persistence." UAT-10147 has been described as a threat actor that conducts search engine op...

Artificial Intelligence (AI) has become one of this decade's defining technologies. From healthcare and finance to manufacturing and education, organizations increasingly rely on AI to automate repetitive tasks, uncover patterns hidden within large datasets, and support faster decision-making. Cybersecurity has experienced a similar transformation. While attackers employ AI to automate cyberattacks and accelerate vulnerability discovery, defenders are adopting AI to improve threat detection and enhance incident response. Security Operations Centers (SOCs) receive a high volume of alerts from endpoints, cloud workloads, network devices, identity providers, and business applications. Although SIEM and XDR platforms provide visibility into these environments, analysts often spend considerable time correlating alerts, searching documentation, and determining the next investigative steps. AI offers a practical way to augment analysts by providing contextual explanations, summarizing ...

A lot of this week’s trouble starts with something trusted doing exactly what it was allowed to do. Signed drivers get turned against defenses. Legitimate apps help malware blend in. A weak header check opens a path to code execution. Elsewhere, exposed systems, old bugs, odd hiding tricks, and AI-assisted exploit research keep lowering the effort needed to cause damage. Nothing here needs much decoration. The small gaps are doing enough work already. The threats change every week. Subscribe, and we’ll alert you when each new ThreatsDay Bulletin is out.

The U.S. government on Wednesday warned of an "active threat" targeting critical infrastructure organizations in the country using artificial intelligence (AI)-generated exploit scripts. The activity is targeting Siemens S7 SeriesProgrammable Logic Controllers (PLCs) to conduct reconnaissance and capability development using AI-generated scripts disguised as legitimate monitoring tools. That said, the ongoing PLC targeting activity is assessed to be broader in scope than Siemens PLCs. "The actors leverage internet scanning services like Censys and ZoomEye to identify internet-exposed PLCs running outdated software or that are otherwise poorly protected," according to the advisory published by the National Security Agency (NSA), Cybersecurity and Infrastructure Security Agency (CISA), Federal Bureau of Investigation (FBI), Department of Energy (DOE), and Environmental Protection Agency (EPA). Targets of the activity include Critical Manufacturing, Energy, Wa...

Adversa AI has disclosed an attack technique that it says can cause xAI's Grok chatbot to send a user's name, approximate location, subscription tier, and the prompts from the ongoing conversation to an attacker-controlled server after the user asks it to summarize an ordinary web page. The AI security company, which has codenamed the technique "Cryptographic Context Injection ," said the transfer completed without a confirmation step and with no visible warning in its proof-of-concept demonstration. There is no patch, no CVE identifier, and no user-facing workaround, and the writeup does not report any exploitation in the wild. Asked which build was tested, Adversa told The Hacker News the target was the Grok web chat at grok.com running Grok 4.5 Fast, and that the attack was reproduced once on August 19, 2026. The writeup gives no success rate. The company said it has attempted the attack 20 times since June with a 40% success rate, and that the failures cam...

OpenAI on Tuesday revealed that it paused reinforcement learning ( RL ) training for its latest artificial intelligence (AI) models for two weeks while it shored up additional defenses and increased the scope of its monitoring to avert another Hugging Face-like incident . "As models become more capable, the risks associated with developing and testing them internally also grow," the AI company said . "Our standards for monitoring, alignment, and security must stay ahead of those risks. We wanted to take the time necessary to meet those standards, so we temporarily slowed the pace of scaling." The company said its largest planned frontier RL run remains on hold for the time being as it conducts smaller-scale training and evaluations to evaluate model behavior, validate its safeguards, and establish more concrete evidence of alignment before moving to the next phase. To that end, OpenAI said it plans to strengthen safeguards across its development process, inc...

A previously unreported cyber espionage operation dubbed SilkParasite has been observed targeting government bodies in Central Asia. The intrusion set makes use of seven remote access tool (RAT) families, five of which have never been previously documented: DriveSilkRAT, CookiETagRAT, NomadRAT, GoginRAT, and NodeEdgeRAT. SilkParasite, first discovered in late 2025, is assessed to be a China-nexus threat cluster with medium confidence. "What makes SilkParasite interesting is the traces of AI-assisted development running through otherwise expert code, which is a different thing from AI-generated malware," Bitdefender Labs said in a technical report shared with The Hacker News. Unlike other operations that rely on AI-generated malware, SilkParasite's arsenal exhibits all hallmarks typically associated with professional espionage tooling that's developed by a team of human operators while AI is likely used to streamline the process. The Romanian cybersecurity v...

Most email defenses still do the job they did a decade ago. Scan the message, look for something malicious, block it. That worked when the danger sat in the payload, a bad link or an attachment. It stopped working when the danger moved into the message's intent, and it is failing now that the sender is no longer a person. From Bad Content to Bad Intent to AI on Both Sides Phishing 1.0 was bad content. Malicious links, infected attachments, spam. Secure email gateways were built for this. Scan the message, match the signature, drop the bad stuff. That era is largely handled. Phishing 2.0 is bad intent. Business email compromise, executive impersonation, fake invoices, wire fraud. There is no malicious payload to scan, only social engineering that reads as a normal request from a person you trust. Gateways are blind to it because there is nothing in the content to flag. Behavioral analysis is the only thing that catches it, which is why some of us have spent the better part ...

Two critical vulnerabilities impacting MLflow, an open-source artificial intelligence (AI) platform, and FUXA, an open-source, web-based SCADA / HMI software built for operational technology (OT) and industrial automation, are witnessing malicious scanning and exploitation efforts. According to independent reports from watchTowr and VulnCheck, the vulnerabilities in question are as follows - CVE-2026-64849 (CVSS score: 9.3) - An unauthenticated Server-Side Request Forgery (SSRF) vulnerability in MLflow that can allow an attacker who can reach the Tracking Server (mlflow server) to issue HTTP requests to arbitrary internal cloud metadata endpoints and extract sensitive data. (Affects versions < 3.15.0) CVE-2026-25895 (CVSS score: 9.5) - A missing authentication for a critical function and path traversal vulnerability in FUXA that can allow an unauthenticated, remote attacker to write arbitrary files to the server file system and achieve remote code execution. (Affects ve...

Security researchers at Anthropic and Switzerland's EPFL have demonstrated that self-propagating payloads can spread from one artificial intelligence (AI) agent to the next through the editable system prompt files that autonomous agent harnesses use to carry state between sessions. The work, released as a preprint on August 10, 2026, tests the technique in a simulated six-agent coding collaboration and in a chain of paired agents modeled on OpenClaw , the open-source autonomous assistant formerly known as Clawdbot and Moltbot . There is no evidence that the technique has spread successfully in the wild, and the same paper reports that a review of archived posts from Moltbook, the social network for AI agents, found no successful agent-to-agent propagation despite several attempts. A one-paragraph warning added to an agent's system prompt reduced spread to near zero across the payloads tested. Fifteen generations of adversarial optimization run against that warning on C...

Cybersecurity researchers at Wiz have disclosed a new GitHub Actions workflow injection vulnerability in Snowflake's public  snowflakedb/snowflake-connector-net repository that it said could be exploited through a crafted GitHub issue to execute commands in a workflow containing internal Jira credentials. The issue was present in  .github/workflows/jira_issue.yml , which ran when a public issue was opened and exposed JIRA_BASE_URL, JIRA_USER_EMAIL, and JIRA_API_TOKEN to the same workflow step. The weakness was confined to the repository's CI/CD automation, with no affected Snowflake Connector for .NET release identified. The workflow inserted attacker-controlled issue title and body values directly into a shell run: block. It also checked github.event.pull_request.user.login even though the event was an issue, meaning the referenced pull request property did not exist. GitHub says , "If you attempt to dereference a nonexistent pr...

Cybersecurity researchers have traced the continued evolution of the Cavern (aka Cav3rn) command-and-control (C2) framework used by Iranian nation-state hackers in attacks targeting entities in Israel. Russian cybersecurity company Kaspersky said its ongoing monitoring of the threat activity cluster since December 2025 has led to the discovery of previously unreported components that expand the toolkit's communication capabilities. "The main finding is a complex C2 module that uses DNS A-record responses to choose between direct HTTPS and a Google Apps Script relay for each transaction," Kaspersky said in an analysis. "The same DNS infrastructure can validate and replace the relay deployment ID, allowing the operator to rotate the Google channel." Cavern, first publicly documented by Check Point Research in early July 2026, consists of multiple moving parts, including an Agent and an assortment of modules, that work in tandem to enable mission-specific...

The expensive attacks are not always the clever ones. This week had plenty of proof. Exposed services got hit, old bugs found fresh use, browser sessions became attack paths, and supply-chain problems kept spreading farther than the original compromise. A lot of it came down to access that was already there and defenses that assumed nobody would look too closely. So, nothing magical. Just a lot of small openings turning into bigger problems. Here’s what stood out. ⚡ Threat of the Week Suspected China APT Behind Exploitation of New VMware Flaw — A suspected China-nexus APT is assessed to be behind the exploitation of a newly patched security flaw in VMware vCenter. The attacks involve the exploitation of CVE-2026-59310 (CVSS score: 9.8), a severe directory-traversal vulnerability in the VMware vCenter server that could be weaponized by a malicious actor to execute arbitrary code. In at least one compromised instance, the attacks led to the deployment of a backdoor and. a rever...

Some weeks have one big security story. Others bring many smaller updates that are easy to miss but still matter. This week has plenty of them, covering cloud services, AI tools, malware, data breaches, scams, and new attack methods. The latest ThreatsDay Bulletin puts all of these short updates in one place, so you can quickly catch up on what happened, what changed, and what security teams should know. The threats change every week. Subscribe, and we’ll alert you when each new ThreatsDay Bulletin is out.

Companies are used to thinking about attackers as outsiders trying to break in. North Korean IT workers flip that model. They apply for jobs, pass interviews, receive legitimate credentials, and can end up inside the same systems companies spend millions trying to protect. That risk is no longer theoretical. The FBI is now investigating a North Korean remote IT worker who reportedly worked for a U.S. federal agency. For CISOs, the priority is clear: spot the warning signs before a fraudulent hire becomes trusted access. When the Threat Gets Hired A recent joint investigation by Mauro Eldritch ( BCA LTD ), Heiner García ( NorthScan ), and ANY.RUN showed what this looks like from inside the operation. Researchers deliberately hired suspected DPRK developers linked to Lazarus Group and gave them what looked like ordinary virtual desktops. In reality, they were controlled ANY.RUN Sandboxes, capturing their activity in real time. The operation exposed forged identities, re...

A newly disclosed flaw in the way OpenAI, Anthropic, and Google carried hidden AI reasoning between API calls let researchers recover internal reasoning and secrets from session logs, including API keys and passwords. The weakness affected encrypted reasoning objects used by the providers' reasoning APIs, where a block created in one session could be replayed into another and, during testing, even handed to a weaker model in the same provider family to make it reveal the hidden content. The team behind the paper  Stealing Reasoning Traces from Proprietary LLM APIs  demonstrated four abuse paths: stealing proprietary reasoning for  model distillation , extracting private data from other users' published traces, recovering harmful content concealed behind a safe visible answer, and hiding prompt injections inside opaque reasoning blocks. Across 6,708 public agent trajectories, the team decoded 315,320 thinking blocks. After excluding benchmark sources, it counted 704...

Anyone sharing their screen on a Zoom call could have taken over the computers of everyone watching, and anyone watching could have taken over the presenter's. The flaw sat in the annotation tool, the feature that lets participants draw and type on a shared screen, and it asked nothing of the victim beyond being in the meeting. No click, no download, no prompt, and nothing on screen to show it had happened. The patches are not new. Client fixes shipped in June and July, roughly two months before the flaws were made public, and no exploitation has been reported as of publication. None of the three identifiers appear in CISA's Known Exploited Vulnerabilities catalog. The versions that close them: Zoom Workplace, all supported platforms, before 7.1.5 and 7.0.6 in their respective branches Zoom Workplace VDI Client for Windows, before 7.0.11 and 6.6.16 Zoom Rooms and Zoom Meeting SDK, all platforms, before 7.1.0, and before 7.1.5 for the third flaw The research ...

Security researchers found a way to enter Microsoft SharePoint servers as any user, including an administrator, with no valid account. A significant part of the work that found it was done through an AI agent. The flaw, tracked as  CVE-2026-55040  (CVSS 9.1), affects SharePoint Server Subscription Edition, SharePoint Server 2019, and SharePoint Server 2016. Microsoft's affected-product list covers only those three on-premises editions, and SharePoint Online is not among them. It lets a remote unauthenticated attacker assume a chosen user's identity. The attack has one prerequisite: the intruder has to know which account they want to become, either by its Active Directory security identifier (SID) or its user principal name (UPN), which is formatted like an email address. Rapid7 then chained the bypass to a separate remote code execution flaw and ran code on the server with no credentials. Microsoft and the firm disclosed that second flaw on August 11 as  CVE-2026...