OpenAI banned Russian ChatGPT accounts backing a fake think tank, IBI, that used AI posts and a fake “sovereignty” index to push pro‑Russia narratives. OpenAI says it has banned a cluster of ChatGPT accounts that likely originated in Russia and were used to support a covert influence operation. The campaign promoted an organisation called the International Burke Institute, or IBI, a supposed expert community that claimed to be based in Israel. The operation did not rely on dramatic deepfakes or a viral bot army. It used a more familiar method: build something that looks credible, fill it with enough content to seem legitimate, then use social media to push people toward it. Credibility is often cheaper to fake than to earn. OpenAI says the operators prompted ChatGPT in Russian to generate comments and posts, mostly in English, for X, Facebook, LinkedIn, Telegram and Substack. They specifically asked the model to avoid linguistic signals that might reveal a Russian origin, and used VPNs to bypass OpenAI’s restriction on access from Russia. “We banned a cluster of ChatGPT accounts that very likely originated in Russia. The operators prompted in Russian to generate social media comments that were posted on Substack, Telegram, X, Facebook and LinkedIn.” reads the report published by OpenAI, “Most of the comments they generated were in English, and the operators instructe. ChatGPT to hide any linguistic clues that they were Russian. As we do not allow access to our models from Russia, they used VPNs to access our platform.” The social media content promoted IBI articles or urged readers to follow IBI-linked channels. Some posts came from accounts bearing the institute’s name and logo, while others appeared to come from ordinary users whose main activity was sharing the same material. The website behind the brand was registered in February 2025 and presented itself as an Israel-based community of international experts. It claimed links to well-known names such as Francis Fukuyama and Noam Chomsky, but OpenAI’s review found that 34 of 36 sampled articles published under expert profiles had been copied from elsewhere online. “Some of these articles were years old; others were attributed to the wrong authors. For example, one article on the China-Pakistan Economic Corridor appears to have been copied from a Cambridge University Press original, but incorrectly attributed to a professor at the University of Nottingham whose expertise is in South Asian politics.” continues the report. This was not a case of ChatGPT writing every false article on the site. OpenAI says the website content itself was not generated by its models, and some material appeared to have been written by a Slavic-language speaker and machine-translated. The AI’s role was narrower but still useful: it helped create the promotional layer that sent people toward the site and made the operation appear more active and organic. openai “What began as an investigation into AI-generated social media posts led us to a much broader influence operation, built around a website containing copied and misattributed academic work, a “sovereignty” index that cast Russia in a favourable light, and efforts to disguise the operators’ Russian origins.” states OpenAI. “Although the campaign appears to have reached relatively small audiences, its elaborate construction distinguishes it from other Russia-linked(opens in a new window) influence(opens in a new window) operations(opens in a new window) we have disrupted(opens in a new window) since the start of the war in Ukraine.” The centrepiece was the so-called Sovereignty Index, also called the Burke Index. It ranked countries across political, economic, technological, information, cultural, cognitive and military dimensions, and consistently gave Russia a favourable place while criticising Western states, especially France, Germany, the European Union and the United States. openai A made-up index can be effective because numbers carry authority even when the method is vague, selective or impossible to audit. Add charts, expert profiles, academic-looking articles and a professional website, and a claim can travel much further than a Telegram post from an anonymous account. OpenAI found one Telegram channel, “Lahme Ente,” that published German-language posts attacking Ukraine, the EU and the German government while calling for closer ties with Russia. Another operator used ChatGPT to create logos for channels focused on Germany, France, Poland, Türkiye and the United States, then repeatedly asked for Russian-language summaries of their activity. “As well as generating content about IBI, one of the operators generated German-language posts that were posted on a Telegram channel called “Lahme Ente” (“lame duck”). These posts routinely criticized Ukraine, the EU and the German government, and advocated for better relations with Russia.” continues the report. “A second operator, alongside their IBI-related content, generated logos for a dozen Telegram channels (including Lahme Ente) focused on Germany, the USA, France, Poland and Türkiye. “ One American-facing channel called “American Observer” included awkward English that suggested it was not run by a native speaker. The operators were trying to hide their origin, but language remains a stubborn problem: a VPN can change an IP address, not always a sentence. The campaign’s immediate reach appears limited. OpenAI says most social posts received few views and the official IBI accounts had low subscriber counts, although its Telegram channels reportedly attracted around 10,000 to 20,000 followers each. Using the Brookings Breakout Scale, OpenAI rated the effort at the lower end of Category Three: activity across multiple platforms with some signs of reaching genuine audiences. That does not make it irrelevant. Influence operations are often built for scale, reuse and timing rather than instant virality. A network with a website, a recognisable brand, social accounts, apparent experts and ready-made narratives can remain quiet until a political event, election, protest or crisis gives it an opening. This case also shows what AI changes and what it does not. It does not eliminate the need for operators, infrastructure, stolen material, audience research or political intent. It makes routine tasks cheaper: drafting posts, translating them, changing tone, generating replies, creating branding and monitoring channels at a pace that a small team can sustain. “The significance of the operation lies less in the audience it reached, however, than in the infrastructure it had built. While the actors only used ChatGPT to produce isolated promotional posts, those posts pointed to an otherwise credible-appearing institution, complete with purported experts, republished academic work and a purported proprietary risk index.” concludes the report. “This illustrates how influence actors can use AI as a supporting tool within a broader effort to manufacture authority, obscure the source of favored narratives, and establish assets that could be scaled over time. It also illustrates how their supporting use of AI can lead to the broader operation being exposed.” Follow me on Twitter: @securityaffairs and Facebook and Mastodon Pierluigi Paganini (SecurityAffairs – hacking, OpenAI)
CISA red teams fully compromised two critical infrastructure orgs. One SOC isolated hosts in minutes; the other never detected the breach. CISA published an advisory (AA26-237A) documenting two simultaneous red team assessments at critical infrastructure organizations. Both organizations lost full domain control and had their cloud environments compromised. One of them didn’t know until CISA told them afterward. “The Cybersecurity and Infrastructure Security Agency (CISA) conducted simultaneous red team assessments at two organizations and observed different defensive outcomes. In both environments, the red team achieved full domain compromise and accessed sensitive business systems (SBSs) and cloud resources.” states CISA. “Organization A failed to detect or contain the activity, but Organization B rapidly identified initial compromise attempts, isolated affected systems, and forced the red team into an assume breach model.” Organization A is a Government Services and Facilities Sector entity. Organization B operates in the Water and Wastewater Systems Sector. The red team used comparable techniques against both. The difference in outcome was entirely about detection and response, not the sophistication of the attack. At Organization A, the red team found a web application that still used default credentials. They used it to send phishing emails from a trusted internal address and gained access to four workstations. From there, they exploited a misconfigured Active Directory Certificate Services template with the ESC1 flaw. This allowed a low-privileged user to request certificates for other users, including administrators. They then reached all the targeted sensitive business systems without anyone noticing. After moving into the cloud, they even read SOC staff emails to see if the attack had been detected. It hadn’t. “Without well-defined baselines and alert filtering, false positives and routine alerts overwhelm defenders, obscuring real threats.” CISA continues. “Organizations that tune alerts to highlight anomalies and filter out normal business activity enable defenders to focus on genuine incidents and respond rapidly.” Organization A’s SOC was receiving thousands of false positive alerts, many at higher severity than the actual intrusion alerts the red team was generating. Staff eventually reviewed SCCM-related alerts from real red team activity, couldn’t identify the system’s owner or function, and marked it a false positive. The red team confirmed the miss by reading SOC email. Then they used keyloggers and screenshot capture on SOC workstations to make sure nothing was coming. Nothing was. The organization had multiple separate SOCs with different EDR solutions and no cross-team visibility, which meant that even if one team noticed something, there was no mechanism to act on it across the relevant systems. “Detection tools are only as effective as the people, processes, and procedures supporting them. SOC staff should not operate in silos and should have clear authority unhindered by bureaucracy to effectively contain and resolve incidents.” add CISA. At Organization A, SOC analysts were managing systems they didn’t fully understand and had no written escalation procedures, so their default response to ambiguity was to wait. At Organization B, staff triaged, investigated, coordinated with engineering, and reimaged machines before handing them back to users. At Organization B, the red team still found important security gaps. They discovered a password stored in plain text inside an XML file on an SCCM distribution point. They used the related service account to gain powerful rights over a domain controller and then performed a DCSync attack, obtaining the krbtgt hash. This allowed them to create Golden Tickets and impersonate users across the domain. They also found a path into the OT network through RDP files pointing to a bastion host. Using FTP credentials found on a jump server, they connected to the bastion through SSH. The bastion had no outbound internet access, so their payload could not run, and the SOC quarantined the host. Still, the access path was there. Both organizations also had the same cloud security problem: neither had enabled Conditional Access for workload identities. This Microsoft feature applies access controls to applications and service accounts, not just human users. Without it, applications with broad Microsoft Graph permissions can bypass normal Conditional Access rules. CISA’s red team used this gap in both organizations to access emails across the companies. In Organization A, the team also found AWS IAM credentials stored in users’ home directories with no expiration date. Those credentials could remain valid indefinitely, creating another long-term risk. In Organization B’s cloud environment, the red team abused Seamless SSO by using Kerberos tickets obtained via DCSync to authenticate to Azure without needing any user’s cleartext password. They found a disabled AD-synced account that owned an application with permission to read, write, and send emails for every user in the tenant. They re-enabled the account, DCSynced its credentials, added a client secret to the application, and could then access the full mailbox of every employee from the public internet. Organization B’s detections flagged the AzureHound tool by user agent and caught anomalous Microsoft Graph API request volumes, but those controls arrived after the initial cloud access was already established. CISA recommends several practical steps to improve security. These include hardening ADCS by disabling CT_FLAG_ENROLLEE_SUPPLIES_SUBJECT on templates and limiting who can enroll, setting the Machine Account Quota to zero when there is no operational need, and removing cleartext credentials from workstations and network shares. Organizations should also enable Conditional Access for workload identities, create procedures to revoke tokens, and treat SCCM and similar endpoint management platforms as Tier 0 assets, giving them the same level of protection as domain controllers. The full advisory also maps each red team technique to its MITRE ATT&CK identifier and compares how well the two organizations detected the different stages of the attacks. Follow me on Twitter: @securityaffairs and Facebook and Mastodon Pierluigi Paganini (SecurityAffairs – hacking, CISA)
A newly disclosed Rowhammer attack called GPUThor can bypass error-correcting code (ECC) protections on NVIDIA GPUs, enabling denial-of-service (DoS) and root-level privilege escalation. In a paper published by the University of Toronto, researchers say that GPUThor achieves far more practical bit-flip rates than past concepts like their own GPUHammer or GPUBreach, which became irrelevant after ECC was introduced. The attack was demonstrated against Ampere-class NVIDIA workstation GPUs with GDDR6 memory, including the RTX A4000, RTX A4500, RTX A5000, and RTX A6000, all widely used in AI and cloud infrastructure. Rowhammer is the name for a class of attacks where memory rows are repeatedly accessed (“hammered”) in a way that increases the likelihood of bits in neighboring memory regions to flip, changing their state from one to zero or vice versa.
FBI seizes QScan and QTRouter, China-linked platforms used to hide intrusions and target U.S. critical infrastructure. The U.S. Department of Justice and the FBI have seized two platforms, QScan and QTRouter, used by a China-linked group to hide cyberattacks and target critical infrastructure. The operation matters because it shows how state-backed actors no longer need to build every part of an attack from scratch: they can rely on shared scanning tools, compromised devices, commercial proxy services, and rented servers. U.S. authorities attribute the activity to a group known as QTFY, which they say operated through the China-based company Nanjing Xinjiuwei Network Technology Company. According to the Justice Department, the group offered hacking services to paying clients that included China’s Ministry of State Security and the People’s Liberation Army. “The Justice Department and FBI announced court-authorized domain seizures today to deny malicious cyber actors access to two complementary hacking platforms known as “QScan” and “QTRouter,” used to target U.S. critical infrastructure and other sensitive networks. As described in court documents unsealed in the Southern District of California, a People’s Republic of China (PRC) state-sponsored group known as “QTFY,” employed by China-based Nanjing Xinjiuwei Network Technology Company (南京鑫玖维网络科技有限公司), created and operated QScan and QTRouter.” reads the press release published by DoJ. According to the Justice Department, QTFY targeted NASA, the Federal Reserve, the Departments of Energy, Justice, and Health and Human Services, the National Institutes of Health, and the U.S. Senate. Those are not random targets caught in a broad scan. QScan and QTRouter had different jobs. QScan searched the internet for vulnerable devices and automatically infected thousands of exposed Internet of Things devices. Those compromised routers, cameras, appliances, and other connected systems were then added to QTRouter, a proxy network used to route malicious traffic through systems located outside China. That approach makes attribution and blocking harder. A victim may see an attack coming from what appears to be a local router, a legitimate commercial proxy, or a virtual private server rented in another country. Blocking traffic by country or IP address becomes far less useful when the attacker is borrowing the identity and location of other people’s devices. “QScan scans and automatically infects thousands of “internet-of-things” (IoT) devices worldwide, which are then added to the QTRouter network of QTFY-controlled devices. QTRouter consists of these compromised IoT devices, as well as commercial proxy service devices and leased virtual private servers.” continues DoJ. “QTRouter then serves as an “obfuscation network” – meaning it allows QTFY and other malicious cyber actors to conceal the PRC-origin of their computer intrusion activities because the malicious communications appear to originate from computers (such as those compromised by QScan) that are outside of the PRC and may even be local to the targeted networks.” The government did not merely take down a website. The seized domains were hard-coded into QScan and QTRouter and used for core functions such as authentication and command-and-control communication. By taking control of those domains through a court order, the FBI made the malware platforms unable to operate as intended. This is a useful example of technical disruption rather than simple attribution. Publicly naming a group matters, but it does not stop an operator who already has access to vulnerable devices and a working relay network. Taking away the infrastructure they need to manage that network is more disruptive, at least until they rebuild it. “Because the seized domains were hard-coded into both the QScan and QTRouter malware and used for essential tasks such as communication and authentication, the court-authorized seizures made QScan and QTRouter inoperable.” DoJ states. The broader operation looks like an industrial service model. QTFY allegedly used QScan to map targets, identify weaknesses, and compromise devices; it then used QTRouter and other proxy layers to hide where the activity actually came from. Lumen’s Black Lotus Labs described the group as a “digital quartermaster,” meaning it supplied infrastructure and services that could support multiple operations at the same time. “To support its primary objectives, the quartermaster runs a secondary, completely decoupled target profiling utility called “QScan.” While the core proxy network focuses on managing stateful session paths, the QScan framework operates as the front-end scout.” reads a report published by Lumen. “The operations of this quartermaster demonstrate the high degree of industrialization occurring within China-nexus cyber operations. By shifting away from fragmented, ad hoc setups and toward shared multi-tenant utility networks, state-sponsored actors can execute complex campaigns with a high degree of anonymity and speed, and at a global scale.” The group reportedly exploited both old and newly disclosed vulnerabilities in widely deployed products, including Fortinet SSL-VPN, Citrix ADC, Microsoft Exchange, F5 BIG-IP, Apache Log4j, Atlassian Confluence, Check Point gateways, CrushFTP, Ivanti appliances, and BeyondTrust Remote Support. This is not a reminder that organisations need magic detection. It is a reminder that old vulnerabilities remain useful because somebody, somewhere, will still leave them unpatched. The attack chain also included web shells, remote-access trojans, stolen or legitimate credentials, and compromised IoT devices used as nearby relay points. That combination allows an attacker to blend into normal-looking traffic and avoids the obvious infrastructure that defenders have learned to block. A hostile connection that looks like a local consumer router is still hostile; it just has better camouflage. This seizure follows earlier FBI actions against China-linked botnets. In 2025, the FBI removed PlugX malware from more than 4,000 infected U.S. computers linked to Mustang Panda. In 2024, it disrupted a botnet of hundreds of thousands of IoT devices associated with Flax Typhoon, while in 2023 it acted against a Volt Typhoon botnet used to conceal activity targeting U.S. and foreign critical infrastructure. “For defenders, the operational lesson is straightforward. Inventory internet-facing assets, patch known vulnerabilities quickly, remove unsupported devices, monitor unusual outbound proxy traffic, and do not assume that an IP address tells you who is behind an attack. The same applies to IoT equipment: a router, camera, network appliance, or smart device left exposed and unmanaged can become part of someone else’s attack platform. “This disruption announced today is among a series of court-authorized technical operations against indiscriminate hacking activities by the PRC.” concludes DoJ. Follow me on Twitter: @securityaffairs and Facebook and Mastodon Pierluigi Paganini (SecurityAffairs – hacking, QTRouter)
Meta has reached a proposed settlement worth up to approximately $18 billion with a bipartisan coalition of 52 attorneys generals over allegations that Facebook and Instagram were deliberately designed to encourage compulsive use by children and teenagers.
Attackers are now targeting a chain of two Microsoft SharePoint vulnerabilities that can allow them to execute arbitrary code on unpatched servers, according to threat intelligence company Defused. The first (tracked as CVE-2026-55040) is an authentication bypass flaw in the JWT token validation pipeline that attackers without privileges can exploit to perform operations as a SharePoint site user or administrator. The second (CVE-2026-63520) is a vulnerability in SharePoint's Business Connectivity Services (BCS) that unauthenticated attackers can chain after successfully exploiting CVE-2026-55040 for remote code execution (RCE) on a targeted SharePoint Server. Both flaws have publicly available proof-of-concept (PoC) exploits, released by Rapid7 security researcher Stephen Fewer on August 11 (for CVE-2026-55040, representing the first part of the exploit chain) and by VulnCheck vulnerability researcher Jonathan Peterson on August 24 (for CVE-2026-63520).
The FBI has disrupted infrastructure associated with a technical “quartermaster” that provided reconnaissance, proxy management, and operational routing capabilities for Chinese cyber espionage activities. Black Lotus Labs, the threat research arm of Lumen Technologies, has been tracking the infrastructure for the past year and discovered the components of the framework used in attacks against U.S. critical infrastructure. According to the researchers, the provider offers a reusable service consisting of four distinct operational elements: The infrastructure was used to profile and steal data from U.S. military and defense organizations, government networks, universities and research institutions, aerospace and bioinformatics organizations, healthcare orgs, financial firms, critical infrastructure and energy companies, and enterprise software vendors.
Snowflake is ending password authentication for legacy service accounts, forcing organizations to migrate them to passwordless methods. Token Security explains why the harder challenge is identifying what uses each account, who owns it, and how much access it still needs.
Ubiquiti has released security patches for three new maximum-severity vulnerabilities that threat actors can exploit remotely without privileges. The first (tracked as CVE-2026-77537) lets unauthenticated attackers compromise unpatched devices by exploiting an improper input validation weakness in the UniFi Protect Application video surveillance management platform. Ubiquiti also addressed a CRLF injection flaw (CVE-2026-77550) that remote attackers without privileges can exploit to bypass authentication on UniFi OS devices or instances. "A malicious actor with access to the network could exploit an Improper Neutralization of CRLF Sequences vulnerability found in certain devices running CRLF Injection to bypass authentication to such UniFi OS devices or instances," it explained. The third maximum severity vulnerability patched today is a command injection security flaw (CVE-2026-77554) stemming from improper input validation in the UniFi Talk Application Voice over IP (VoIP) phone system.
Microsoft has begun testing new privacy controls that will let Windows 11 users choose which desktop applications can access their camera, microphone, and precise location. This new feature is shipping to systems upgraded to Windows 11 Insider Experimental Preview Build 26340.9233, and the goal is to make privacy permissions easier to understand in Windows 11 and quicker to manage for each desktop app. "Windows Insiders can now manage camera, microphone, and location permissions for individual desktop apps. Previously, access for traditional desktop applications was managed through a single device-wide setting," Microsoft explained. "With this update, you can review and control access on an app-by-app basis, giving you greater visibility into which apps are requesting access to sensitive resources and more control over your privacy choices."
Confirm this action.
Leaving now will discard your changes.