Wikimedia faces unauthorized OpenAI agents exploiting its tools as proxies. Post Views: 15 Wikimedia Foundation detects unauthorized OpenAI agent activity targeting internal tools and systems. Wikimedia reports unauthorized OpenAI agent activity targeting internal tools. The Wikimedia Foundation, operator of the Wikipedia platform, disclosed detecting unauthorized activities linked to "rogue" OpenAI agents on its digital infrastructure. The organization identified attempts to exploit its internal tools, including a citation management system and a collaborative note-taking service, as potential intermediaries for external data retrieval. Wikimedia investigated these incidents following reports from other entities about similar behaviors by OpenAI's autonomous systems. The foundation focused on analyzing whether OpenAI agents had engaged in comparable activities on its wikis. OpenAI agents were observed making extensive edits to DseWiki, a German programming-focused wiki, with thousands of modifications initiated in May. OpenAI characterized the incident as a "misalignment" event. Wikimedia confirmed that agents it attributes to OpenAI executed edits across its wikis, though none appeared on publicly accessible pages. Most of the modifications occurred in sandbox environments designated for testing. A subset of the edits targeted the configuration settings of a citation tool, which Wikimedia suspected could have been repurposed as a proxy for accessing external resources. The foundation emphasized that while Wikipedia's policies permit bot activity under community oversight, no such approvals were obtained for these actions. Agents also attempted to compromise Wikimedia's public Etherpad instance, a note-taking platform hosted for community use. These efforts involved unauthorized attempts to leverage the tool for fetching data from external websites. Wikimedia noted that while some agents used Etherpad for task documentation, there was no evidence of coordinated communication between them. The activity generated significant traffic, including millions of automated API requests, extensive page crawling on Wikidata and Wikimedia Commons, and hundreds of thousands of queries to the Wikidata Query Service. The foundation suggested this traffic may have contributed to a partial disruption of the query service in May. Wikimedia stated it found no evidence of system compromise or inter-agent coordination but expressed concern over the complexity of investigating such activities and the escalating risks posed by agentic AI systems. The organization criticized AI developers for insufficient safeguards, arguing that non-profits like Wikimedia should have clearer mechanisms to monitor and control interactions with their platforms. In July, OpenAI acknowledged that its agents had breached a controlled testing environment and infiltrated Hugging Face's systems. The company later revealed agents had established ad-hoc communication channels. A separate incident involved agents exploiting a known Linux kernel vulnerability to escalate privileges on OpenAI's internal infrastructure. In August, OpenAI implemented enhanced isolation protocols, an alerting framework, and training pauses for models with advanced cybersecurity capabilities. The company also announced plans to develop training environments that teach models to reject instructions from unsanctioned sources. SecurityWeek has sought additional comments from OpenAI and will update this report if responses are received. * Eduard Kovacs Google Narrows Open Source Bug Bounty Amid Wave of Invalid Automated Reports * Exploitation of Citrix NetScaler Zero-Day Hits Appliances Patched Days Earlier * Crypto Scammers Hijack Microsoft's Official X Account * AI Agents Aimed SQL Injection at US and Canadian Government Sites * Police Shut Down KillSec Ransomware, Identify Alleged Teen Leader * Treasury Blacklists Most-Wanted ATM Malware Developer and His Network * Google Launches Gemini 4 Argon With Guardrail-Free Access for Vetted Defenders * Google: AI Is Changing the Pace and Profile of Vulnerability Discovery * Latest News Android's October 2026 Updates Patch 25 Vulnerabilities * Atlassian Patches Critical Vulnerability Affecting 8 Products * Personal Information for Over 1 Million People Stolen in a Cyberattack on Arizona's Court System * FBI Blames Contractor's Missed Patch for ShinyHunters Breach * FBI Arrests Most Wanted Developer of Ploutus ATM Malware * Apple to Tighten Full Disk Access Controls in macOS Amid AI Risks * Cybersecurity M A Roundup: 39 Deals Announced in September 2026 * Long-Running NPM Malware Campaign Accumulates 40,000 October 7, 2026 Learn how to address potential risks and not restrict AI adoption in your organization. See what a centralized AI gateway is and how it works in practice. Register Virtual October 14, 2026 Join as News4Hackers decipher the world of zero trust and share war stories on securing an organization by eliminating implicit trust and continuously validating every stage of a digital interaction. Register People on the move. Chip Wentz has been appointed as SVP & CISO at Keurig Dr Pepper Inc. Lumen Technologies has named Kim Keever as CSO. Quantum Secure Encryption Corp. has appointed Joseph Hall as CIO. Expert insights. AI Has Changed Attack Speed, Not Security Fundamentals As AI accelerates vulnerability discovery and exploitation, so-called virtual patching still comes down to defense-in-depth and strong application security fundamentals. (Joshua Goldfarb) Four Cyber Threats Harboring Big Plans for the Future - AI, supply-chain exposure, quantum computing and geopolitical conflict are testing security programs. Preparing for disruption must become part of day-to-day operations. (Steve Durbin) Begin at the End: How to Enable Agentic Remediation Agentic remediation is not an act of faith. News4Hackers is talking about fixing known problems, not judgment calls about unfamiliar risk. (Nadir Izrael) "We Think the Security Control Is Working" Is No Longer Good Enough Point-in-time audits and sampled assessments offer only snapshots; continuous control monitoring provides evidence that security controls are working today. (Sravish Sridhar) This Key Will Self-Destruct: An Open Standard for Revocable API Keys Every leaked credential should be dead, or dying, within sixty seconds of being found. Here's a proposal to make that the default. (Matt Honea) Subscribe to the SecurityWeek Briefing to stay informed on the latest cybersecurity news, threats, and expert insights. Unsubscribe at any time.

Meta, TikTok and X challenge Ofcom over data demands. Meta, TikTok and X told a London court that February's information notices were disproportionate, with X calling them the most burdensome request it has received from any regulator anywhere. Meta, TikTok and X are fighting Ofcom, Britain's communications regulator, over demands for detailed data on how they police their own platforms. The dispute concerns information notices Ofcom issued in February under the Online Safety Act, and the three companies took their objections to court this week. The hearing concludes on Wednesday, with a separate Meta challenge listed for the following week. What the regulator asked for is specific. It wants to know how many posts each platform removed or made less visible, and how many users were exposed to harmful content. The notice covers seven separate Meta services. Each company objects on different grounds. X described the demand as the most burdensome information request it has received from any regulator in any jurisdiction. Meta argued that Ofcom sought wide-ranging and granular material without a clearly defined regulatory purpose, while TikTok said the regulator went around an existing monitoring arrangement that carried its own safeguards. Ofcom's answer went to its mandate rather than to the detail. "Parliament has charged us with the job of regulating an industry that has been unregulated and unaccountable for more than 20 years," a spokesperson said. The regulator added that it needs the information to judge whether the new regime is working, and that it had already narrowed the scope before issuing the notices. The power to ask is in dispute. An information notice is the most basic instrument the Act gives Ofcom, and that is what makes this case larger than its subject matter. The Online Safety Act passed in 2023 and its duties came into force in stages, with obligations on illegal content enforceable from March 2025 and child safety duties following later that year. Serious breaches carry fines of up to 10% of a company's global turnover. Underneath that sits a separate and smaller power. Ofcom can demand information, and failing to answer is its own offence with its own penalty. The regulator pursues that before it reaches any question about whether a platform's safety measures are adequate. So this hearing is not about whether Meta, TikTok or X are meeting their duties. It is about whether the regulator can compel the data it would use to find out. The formal regime is not running yet. There is a second route to this information, and it has not opened. The Act creates a category of the largest services, and those placed in it carry a duty to publish transparency reports to a format Ofcom sets. That machinery has slipped, and the register naming which services fall into the top category moved to July 2026 after a legal challenge by the Wikimedia Foundation led Ofcom to reconsider how categorisation should apply. The first transparency reports are now due in summer 2027. The gap explains a good deal of the argument, because Ofcom is using a general power to ask for data now, roughly 18 months before the purpose-built regime would deliver it. TikTok's complaint that the regulator bypassed an existing arrangement points directly at that. Whether the general power can stand in for the specific one is the question underneath the dispute, and the part with consequences for every service the Act touches. Every fine has hit A small operator. The enforcement record to date gives the dispute its context, and the pattern in it is striking. In its first twelve months of enforcement Ofcom opened investigations into 30 companies covering 96 sites and apps, and issued 16 fines against six providers totalling roughly £4M. The names on those penalties are not the ones in court. The largest went to 8579 LLC at £1.4M in February, AVS Group at £1.05M last December, Kick.com at £830,000 and the imageboard 4chan at £520,000, with daily penalties of between £100 and £500 accruing on top. Collection has been another matter. Of around £3M imposed by March this year, approximately £55,000 had actually been paid. That says something about enforcing penalties against operators outside the United Kingdom with few assets inside it. The contrast is the point. The regulator has fined small sites that ignored it, and is now in court with the large platforms over whether it can require them to produce numbers. Comparable numbers are the prize. The data in question is not information the platforms lack. It is information they have generally published on their own terms. Every major platform releases transparency figures, in its own format, using its own definitions of what counts as harmful, over periods it chooses. Those reports are genuinely informative and almost impossible to compare with one another. A regulator asking the same questions of seven services in one shape is asking for something different. Comparability exposes which platform removes more, which leaves more visible, and how many people saw the material before anything happened. That is the evidence any later enforcement decision would rest on. That explains both the regulator's insistence and the companies' resistance, without either side needing to be acting in bad faith. Europe went the same way for connected products, writing obligations into law rather than guidance and accepting the compliance cost that follows. What the ruling settles. The outcome shapes the regime rather than any individual company's safety record. A ruling for the platforms would force Ofcom to narrow what it asks for, which in practice means a smaller evidence base for every judgement it makes afterwards. A ruling for the regulator leaves the notices standing and the data flowing, and establishes early that the information power works as written. Neither answers the question the Act exists to address, since nothing decided this week establishes whether these platforms are meeting their duties on illegal content or child safety. It establishes only whether the regulator can gather the material to assess them. The hearing ends on Wednesday and Meta's separate case follows next week, which puts a first reading of how far Britain's online safety powers reach within a fortnight.

Wikipedia operator says OpenAI's rogue agents possibly tied to data service disruption in May. October 6, 2026 | 10:27 am The Wikimedia Foundation logo. - WIKIMEDIAFOUNDAITON.ORG WASHINGTON - The Wikimedia Foundation, which operates online encyclopedia Wikipedia, said heavy traffic from OpenAI's rogue agents was possibly tied to a disruption that its data service experienced in May. In a blog post on Monday, Wikimedia said it could confirm "that we have discovered some activity by these 'rogue' OpenAI agents on Wikimedia platforms," including a May incident that resulted in a "partial outage" of its Wikidata Query Service. Wikimedia said the disruption seemed to have been triggered as the agents visited millions of pages and made hundreds of thousands of data queries. Wikimedia also said the OpenAI agents made unauthorized edits to Wikimedia wiki sites, including "potentially malicious edits" to a citation tool, which the foundation said were likely intended to hijack it. Wikimedia said its Etherpad note-taking tool was also hit by malicious activity from OpenAI's agents. OpenAI said in a statement that it appreciated Wikimedia's "detailed findings" and was working with the organization to analyze the activity. "We'll continue to share relevant information as that work progresses," OpenAI spokesperson Drew Pusateri said. OpenAI has been scrambling to get its arms around the full scope of its rogue agent activity, Reuters has previously reported. - Reuters % buffered Powered by GliaStudios