Stay updated with the latest news about AI agents, autonomous AI, and automation tools.
NJ1015 warns AI agents acting without human oversight could disrupt NJ's grid and networks, as more towns move to block data centers.
# Every CX start-up funding deal you need to know: September 18 ## The week’s round-up of CX and service start-ups that have announced funding, with details of how much they raised and who backed them - Raindrop raises $35m as it advances AI agent performance monitoring ## **Raindrop raises $35m as it advances AI agent performance monitoring** Raindrop raised $35 million in its recent Series A led by CRV with previous investors Lightspeed Venture Partners, which backed Profound, and Y Combinator also investing along with lead researchers from OpenAI, Anthropic and Thinking Machines. Raindrop monitors production [AI agents](https://www.cxnetwork.com/guides/ai-agents-for-customer-experience-a-complete-guide) for silent failures including costly hallucinations and tool misuse. It's a similar sell to [Temporal Technologies](https://www.cxnetwork.com/cx-experience/news/every-vc-cx-deal-you-need-to-know-august-21), which is also backed by Lightspeed Venture Partners and has a relationship with OpenAI. Llast month Temporal Technologies confirmed it was negotiating a new round of financing that would take its valuation to $12 billion. Simulations replay real production traffic and existing test cases against a proposed change to an agent harness, then apply Raindrop's anomaly detection to the results. Teams can measure performance against their own test cases while also detecting unexpected behavior changes before they reach production. In other words, AI engineers can now see "what their change will change." "Agents now run for hours, call thousands of tools, and handle real money, real health data, and real customers," said CEO Zubin Koticha. Raindrop catches those failures in production, and with Simulations, before a change ever ships. This funding lets us bring the testing process frontier labs use on their own models to every team building agents." ## **Kapture CX closes pre-Series B round with $10m to take custom AI workflows global** It said that in addition t
# Ant International embeds AI agents across all platforms in biggest-ever product upgrade ### The new suite features accounts tailored for AI agents, allowing users to delegate specific tasks without relinquishing account control [**Chasing physical AI: Ant Group robotics arm Robbyant kicks off external funding**](https://www.scmp.com/tech/big-tech/article/3362822/ant-groups-embodied-ai-arm-robbyant-kicks-external-funding?module=topic_widget&pgtype=article) The move follows a roughly US$1.2 billion series A funding round in July, backed by Ant Group, Alibaba Group Holding and other institutional investors. [**Chasing physical AI: Ant Group robotics arm Robbyant kicks off external funding**](https://www.scmp.com/tech/big-tech/article/3362822/ant-groups-embodied-ai-arm-robbyant-kicks-external-funding?module=topic_widget&pgtype=article)
The company said 89.5% of Antom merchants have deployed AI agents over the past year, while 81.4% of payment tasks are now completed with AI assistance. It also supports settlement between AI agents in increments as small as $0.000001.
That takes total funding to $50m, the San Francisco company [said on 17 September](https://www.raindrop.ai/blog/series-a/). ## Who is backing it The company also lists Figma Ventures and Vercel Ventures as backers. ## A category getting crowded Scaled Cognition took $100m from Khosla for [reliable agents](https://thenextweb.com/news/scaled-cognition-100m-khosla-reliable-ai) in June. ## Also tagged with - [Funding](https://thenextweb.com/topic/funding)
Google warns that connecting AI agents to smart home devices can result in 'unexpected or even undesired behavior.'
Are you battling a case of AI anxiety? If so, recent headlines haven't done much to calm the nerves. In July, OpenAI admitted that while conducting a...
OpenAI shared six new examples of AI misalignment. In one case, AI agents taught future versions of themselves to bypass human control.
These systems are generally called AI agents. [The National Institute of Standards and Technology](https://www.nist.gov/news-events/news/2025/08/lessons-learned-consortium-tool-use-agent-systems?utm_source=chatgpt.com) describes the emerging paradigm as placing general-purpose AI models inside software scaffolding that enables them to manipulate tools and take actions beyond simply producing text. Suppose an AI agent has been assigned a difficult objective. The agent continues searching. An AI agent can pursue an objective through a sequence of actions, using tools and observing the results before deciding what to do next. In one experiment, Claude was operating as an email-oversight agent inside a fictitious company. In July 2026, [OpenAI](https://openai.com/es-419/index/hugging-face-incident-and-the-road-ahead/?utm_source=chatgpt.com) was conducting internal cybersecurity evaluations in which AI agents were trying to solve exceptionally difficult security problems. The agents were trying to solve their assigned evaluation problems. OpenAI’s investigation documented agents behaving differently even within the unusual environment that produced the Hugging Face incident, and the company is now developing training environments intended to teach models to remain within their original tasks and permissions even when they encounter new tools, persuasive peers or exposed credentials.
# The fix for rogue AI agents could be more AI 1:34 PM PDT · September 17, 2026 [Share on Facebook](https://www.facebook.com/sharer.php?u=https%3A%2F%2Ftechcrunch.com%2F2026%2F09%2F17%2Fthe-fix-for-rogue-ai-agents-could-be-more-ai%2F)[Share on X](https://twitter.com/intent/tweet?url=https%3A%2F%2Ftechcrunch.com%2F2026%2F09%2F17%2Fthe-fix-for-rogue-ai-agents-could-be-more-ai%2F&text=The+fix+for+rogue+AI+agents+could+be+more+AI&via=techcrunch)[Share on LinkedIn](https://www.linkedin.com/shareArticle?url=https%3A%2F%2Ftechcrunch.com%2F2026%2F09%2F17%2Fthe-fix-for-rogue-ai-agents-could-be-more-ai%2F&title=The+fix+for+rogue+AI+agents+could+be+more+AI&summary=As+companies+hand+off+longer+and+more+complex+tasks+to+AI+agents%2C+they+are+running+into+an+oversight+problem%3A+Agents+can+act+faster%2C+longer%2C+and+at+greater+volume+than+humans+can+realistically+review.+&mini=1&source=TechCrunch)[Share on Reddit](https://www.reddit.com/submit?url=https%3A%2F%2Ftechcrunch.com%2F2026%2F09%2F17%2Fthe-fix-for-rogue-ai-agents-could-be-more-ai%2F&title=The+fix+for+rogue+AI+agents+could+be+more+AI)[Share over Email](mailto:?subject=The+fix+for+rogue+AI+agents+could+be+more+AI&body=Article%3A+https%3A%2F%2Ftechcrunch.com%2F2026%2F09%2F17%2Fthe-fix-for-rogue-ai-agents-could-be-more-ai%2F)[Copy Share Link](https://techcrunch.com/2026/09/17/the-fix-for-rogue-ai-agents-could-be-more-ai/) That issue reached a peak with the Hugging Face incident, which saw nearly 12,000 agents coordinating faster than human beings could track. “If you’ve got an AI that’s doing malicious things and it suspects that another AI is keeping tabs on it, it could try and trick that AI,” said Simon Willison, influential tech blogger who has tracked a string of AI agent incidents this year. “We saw a little bit of this in the Hugging Face incident with OpenAI, where their models were all conspiring together to trick a grading AI so that they could get illicit answers past the thing. 00:00 After the July Hugging
# [Inside Precision Medicine](https://www.insideprecisionmedicine.com/) [Inside Precision Medicine](https://www.insideprecisionmedicine.com/)[Informatics](https://www.insideprecisionmedicine.com/category/topics/informatics/) AI Agents Collaborate to Streamline Drug Discovery An AI system that coordinates computerized “scientist” agents under a virtual chief scientific officer could speed the discovery of new therapeutic drugs. Looking across 55,984 clinical trials, more than 37,000 AI agents found that drugs targeting genes specific to particular cell types were 48% more likely to reach the market and were associated with 32% fewer adverse events.
A.I. is already a powerful tool for scientific discovery. Now, some biomedical researchers are teaching it to do even more.
## Virtual biotech company puts thousands of AI scientist agents to work on drug discovery By [Hanae Armitage](https://med.stanford.edu/news/media-contacts/hanae_armitage.html) An artificial intelligence-backed virtual biotech company predicted drug trial success and independently designed a lung cancer therapy that was later validated in trials. ##### Share Article: [Facebook](https://www.facebook.com/sharer/sharer.php) [LinkedIn](https://www.linkedin.com/sharing/share-offsite/) [Twitter](https://twitter.com/intent/tweet) [Email](mailto:) The latest company to spin out of a Stanford Medicine lab is a biotech undertaking with 37,000 employees — and none of them are human. It’s an [artificial intelligence](https://med.stanford.edu/news/topics/artificial-intelligence.html)-powered virtual biotech company that’s the brainchild of associate professor of biomedical data science [James Zou](https://profiles.stanford.edu/james-zou), PhD, who is also the principal investigator of a [virtual lab](https://med.stanford.edu/news/all-news/2025/07/virtual-scientist.html) that launched in 2025, and graduate student Harrison Zhang. They created an entire company with tens of thousands of AI agents, all trained to support the full pipeline of drug development. Zou said. “Could we have a fully agentic company that tackles the extremely complex challenges of drug discovery?”  James Zou Zou’s virtual company more or less mirrors the organizational chart of an established brick-and-mortar biotech: A chief science officer agent leads the research teams, which are broken into multiple specialized divisions that work in parallel to tackle the core elements of drug design, such as identifying molecular targets and designing clinical trials. Zou is the senior author, and Zhang is the lead author. ### A
OpenAI’s agents broke out of their testing environment “sandbox” in July, gained access to the internet, and compromised parts of the systems of Hugging Face, an AI development hub, according to [OpenAI’s August 26 account](https://openai.com/index/hugging-face-incident-and-the-road-ahead/).
## Share this: This year, AI agents have been escaping their confines to go on hacking sprees. In July, news broke that AI agents had snuck out of what was supposed to be an isolated test environment at OpenAI. The agents communicated on a secret message board as they breached private systems at the company Hugging Face, searching for answers to the test they were taking. Most recently, news has emerged that during other tests at OpenAI from this past spring, agents that were only supposed to be looking and not touching anything online wound up messaging each other on at least [ten different online message boards](https://www.reuters.com/world/openais-rogue-agents-used-least-10-more-sites-unauthorized-comms-researchers-say-2026-09-09/). ## Who let the dogs out? In April, several months before the “rogue AI” uproar, an AI agent that was supposed to be checking whether a test version of his company’s software matched the live version instead wound up [deleting everything.](https://x.com/lifeofjer/status/2048103471019434248?s=61&utm_source=substack&utm_medium=email) It wiped out the company’s live data and all their most recent backups. With the way many AI agents today are set up, “you have decisions that are made in split seconds without a human in the loop, whether good or bad,” Crane says. ## More Stories from Science News on [Artificial Intelligence](https://www.sciencenews.org/topic/artificial-intelligence) ### [AI agents aren’t ready to replace humans in behavioral research](https://www.sciencenews.org/article/ai-agents-replace-humans-research) By [Sujata Gupta](https://www.sciencenews.org/author/sujata-gupta)September 3, 2026
## Trending News ### Eagle Cloud Closes Series B+ Round to Expand AI Agent Governance Platform and International Operations [](https://d2c0db5b8fb27c1c9887-9b32efc83a6b298bb22e7a1df0837426.ssl.cf2.rackcdn.com/28536159-eagle-cloud-closes-series-b-fun-2400x1185.png#2400x1185) Eagle Cloud closes series B funding. _Round co-invested by MTR Lab and Northern Light Venture Capital follows the company's Series B by four months_ HONG KONG, HONG KONG, September 17, 2026 / [EINPresswire.com](https://www.einpresswire.com/)/ -- [Eagle Cloud](http://www.eaglecloud.com/en), a provider of AI security governance infrastructure, has [closed a Series B+ funding](https://lnkd.in/p/dW87xr48) round of close to RMB 100 million (approximately USD 14 million), four months after completing its Series B. The round was co-invested by [MTR Lab](https://www.mtrlab.com.hk/en/news/eagle-cloud-raises-series-b-plus-backed-by-mtr-lab/) and Northern Light Venture Capital, with Voyagers Partners acting as financial advisor. The company said the funding will be used to extend its governance infrastructure to AI agents and other forms of autonomous action inside the enterprise, and to accelerate international expansion through a larger overseas team and stronger local delivery capability. "Through this investment, MTR Lab hopes to support Eagle Cloud in further enhancing its technology solutions and expanding its overseas service capabilities."
# Barcelona-based Health Force raises €4.2 million to streamline hospital operations with AI agents - [AI Agents](https://www.eu-startups.com/tag/ai-agents/) - [Health Force](https://www.eu-startups.com/tag/health-force/) - [healthtech](https://www.eu-startups.com/tag/healthtech/) - [Jorge Blasco](https://www.eu-startups.com/tag/jorge-blasco/) - [LUMO labs](https://www.eu-startups.com/tag/lumo-labs/) ### [Málaga-based Atomic One raises €5.6 million to automate e-commerce operations with AI agents](https://www.eu-startups.com/2026/07/malaga-based-atomic-one-raises-e5-6-million-to-automate-e-commerce-operations-with-ai-agents/) July 31, 2026 [Germany-Startups](https://www.eu-startups.com/category/germany-startups/) [YC-backed telli raises €13.1 million to build AI for B2C customer operations](https://www.eu-startups.com/2026/07/yc-backed-telli-raises-e13-1-million-to-build-ai-for-b2c-customer-operations/) ### [YC-backed telli raises €13.1 million to build AI for B2C customer operations](https://www.eu-startups.com/2026/07/yc-backed-telli-raises-e13-1-million-to-build-ai-for-b2c-customer-operations/) [Spain-Startups](https://www.eu-startups.com/category/spain-startups/) [Barcelona-based NeuralTrust raises €17.2 million to secure and govern enterprise AI agents](https://www.eu-startups.com/2026/06/barcelona-based-neuraltrust-raises-e17-2-million-to-secure-and-govern-enterprise-ai-agents/) ### [Barcelona-based NeuralTrust raises €17.2 million to secure and govern enterprise AI agents](https://www.eu-startups.com/2026/06/barcelona-based-neuraltrust-raises-e17-2-million-to-secure-and-govern-enterprise-ai-agents/) June 17, 2026 [Polish FinTech paymove secures €2.12 million to drive European expansion](https://www.eu-startups.com/2026/06/polish-fintech-paymove-secures-e2-12-million-to-drive-european-expansion/) ### [Oslo-based Cloudgeni raises €858k to build reliable AI agents for secure cloud infrastructure](https://www.eu-startups.com/2026/05/oslo-based-cloudgeni-r
The investment was co-led by existing shareholders Time Folio Asset Management and new investor Stone Bridge Ventures. IMM Investment and Stick Ventures...
OpenAI has disclosed six reports on unexpected or concerning behavior in artificial-intelligence models. 1:36. Headlines from ABC News LiveCatch up on the...
TechCrunch Disrupt 2026 in San Francisco will host a session on building startup teams where people work alongside AI agents.
The next member of a startup team may not be a person, but an AI agent. These systems are already handling engineering, research, and operational tasks,...
# OpenAI launches a new framework to track and investigate rogue AI agents OpenAI launched a framework for publicly reporting model misalignment and disclosed more incidents of rogue agents.Benjamin Fanjoy/Getty Images Sep 16, 2026, 7:43 PM ET [Read in app](https://insider-app.onelink.me/4cpG/?af_js_web=true&af_ss_ver=2_3_0&af_dp=insider%3A%2F%2Fbi%2Fpost%2Fopenai-unveils-a-system-for-reporting-rogue-ai-agent-behavior-2026-9&af_force_deeplink=true&is_retargeting=true&deep_link_value=https%3A%2F%2Fwww.businessinsider.com%2Fopenai-unveils-a-system-for-reporting-rogue-ai-agent-behavior-2026-9&pid=businessinsider&c=post_page_share_bar_v2_smart_4.13.23) OpenAI is putting its [misbehaving models](https://www.businessinsider.com/ai-agents-rogue-strategies-cheating-lying-german-wiki-openai-anthropic-2026-9) on the record. The AI company announced a new framework on Wednesday for tracking, investigating, and publicly disclosing cases of model misalignment, alongside six reports detailing concerning behavior observed during training or evaluation over the past six months. During the training of GPT-5.6 Sol, models similarly left themselves instructions to conceal mistakes. OpenAI said that other agents had also searched public repositories for exposed API keys, uploaded files to the internet so they could cite them, and used an internal software repository to communicate across separate training samples. The framework follows an incident in which an OpenAI model escaped a research sandbox and accessed Hugging Face's production systems while operating with reduced safeguards.
# Enterprise Agentic AI Firm Enhans Raises USD 39M Series C, Bringing Total Funding to USD 61.5M 9/17/2026 ・ [sjyoon](https://en.wowtale.net/author/sjyoon/) [](https://www.facebook.com/sharer.php?u=https%3A%2F%2Fen.wowtale.net%2F2026%2F09%2F17%2F235141%2F)[](https://twitter.com/share?url=https%3A%2F%2Fen.wowtale.net%2F2026%2F09%2F17%2F235141%2F&text=Enterprise+Agentic+AI+Firm+Enhans+Raises+USD+39M+Series+C%2C+Bringing+Total+Funding+to+USD+61.5M)[](https://line.me/R/msg/text/?Enterprise%20Agentic%20AI%20Firm%20Enhans%20Raises%20USD%2039M%20Series%20C,%20Bringing%20Total%20Funding%20to%20USD%2061.5M%0D%0Ahttps://en.wowtale.net/2026/09/17/235141/)[](https://www.band.us/plugin/share?body=Enterprise+Agentic+AI+Firm+Enhans+Raises+USD+39M+Series+C%2C+Bringing+Total+Funding+to+USD+61.5M%20|%20https%3A%2F%2Fen.wowtale.net%2F2026%2F09%2F17%2F235141%2F)[](mailto:?subject=Enterprise%20Agentic%20AI%20Firm%20Enhans%20Raises%20USD%2039M%20Series%20C,%20Bringing%20Total%20Funding%20to%20USD%2061.5M&body=https%3A%2F%2Fen.wowtale.net%2F2026%2F09%2F17%2F235141%2F%0D%0A%0D%0A--%20Sent%20via%20PAVO%20SocialSharing)[](https://en.wowtale.net/202
# Manus Reportedly Finalizing $500 Million Funding Round, Valuation Doubles to $4 Billion Chinese AI startup Manus is reportedly close to completing a $500 million funding round that would double its valuation to $4 billion, making it the most valuable startup in China's AI agent sector. ##### Key Elements Chinese artificial intelligence startup Manus is reportedly on the verge of completing a $500 million funding round, which would double the company's valuation to $4 billion and catapult it to become the most valuable startup in China's AI agent sector.
[Skip to content](https://techcrunch.com/2026/09/16/your-startups-next-teammate-might-be-an-ai-agent-gusto-insight-partners-and-leland-explain-what-that-changes-at-techcrunch-disrupt-2026/#wp--skip-link--target) [Share on Facebook](https://www.facebook.com/sharer.php?u=https%3A%2F%2Ftechcrunch.com%2F2026%2F09%2F16%2Fyour-startups-next-teammate-might-be-an-ai-agent-gusto-insight-partners-and-leland-explain-what-that-changes-at-techcrunch-disrupt-2026%2F)[Share on X](https://twitter.com/intent/tweet?url=https%3A%2F%2Ftechcrunch.com%2F2026%2F09%2F16%2Fyour-startups-next-teammate-might-be-an-ai-agent-gusto-insight-partners-and-leland-explain-what-that-changes-at-techcrunch-disrupt-2026%2F&text=Your+startup%E2%80%99s+next+teammate+might+be+an+AI+agent%3A+Gusto%2C+Insight+Partners%2C+and+Leland+explain+what+that+changes+at+TechCrunch+Disrupt+2026&via=techcrunch)[Share on LinkedIn](https://www.linkedin.com/shareArticle?url=https%3A%2F%2Ftechcrunch.com%2F2026%2F09%2F16%2Fyour-startups-next-teammate-might-be-an-ai-agent-gusto-insight-partners-and-leland-explain-what-that-changes-at-techcrunch-disrupt-2026%2F&title=Your+startup%E2%80%99s+next+teammate+might+be+an+AI+agent%3A+Gusto%2C+Insight+Partners%2C+and+Leland+explain+what+that+changes+at+TechCrunch+Disrupt+2026&summary=This+session+will+explore+how+early-stage+companies+are+building+teams+where+humans+and+AI+agents+work+alongside+each+other+%E2%80%94+and+how+founders+can+do+that+without+sacrificing+speed%2C+accountability%2C+or+culture.+Learn+more+at+TechCrunch+Disrupt+2026.+Register+before+September+25+to+save+up+to+%24200.&mini=1&source=TechCrunch)[Share on Reddit](https://www.reddit.com/submit?url=https%3A%2F%2Ftechcrunch.com%2F2026%2F09%2F16%2Fyour-startups-next-teammate-might-be-an-ai-agent-gusto-insight-partners-and-leland-explain-what-that-changes-at-techcrunch-disrupt-2026%2F&title=Your+startup%E2%80%99s+next+teammate+might+be+an+AI+agent%3A+Gusto%2C+Insight+Partners%2C+and+Leland+explain+what+that+changes+at+Te
# How to Use AI Agents to Prepare 3D Scenes for Simulation ## AI-Generated Summary ### Next Steps Specialized subagents built with the Hermes agent harness and deployed through [NVIDIA NemoClaw](https://www.nvidia.com/en-us/ai/nemoclaw/?ncid=pa-so-yout-961923&_bt=804567865336&_bk=nvidia%20nemoclaw&_bm=e&_bn=g&_bg=197993095849&gad_source=1&gad_campaignid=23744621431&gbraid=0AAAAAD4XAoF_WgK5gcBBurrhExtZ8MqXP&gclid=CjwKCAjwmJjSBhB-EiwAkZgxi5RGpiph8SgXKuqkPtDk3e8BO1V1wa5IhxAYWy7hj_XabVunIx_T1RoCTtwQAvD_BwE) use Omniverse Libraries to inspect the scene, author simulation metadata, configure physics, and render visual preflight views. ## Tags [Agentic AI / Generative AI](https://developer.nvidia.com/blog/category/generative-ai/) | [Simulation / Modeling / Design](https://developer.nvidia.com/blog/category/simulation-modeling-design/) | [General](https://developer.nvidia.com/blog/recent-posts/?industry=General) | [Isaac Lab](https://developer.nvidia.com/blog/recent-posts/?products=Isaac+Lab) | [Isaac Sim](https://developer.nvidia.com/blog/recent-posts/?products=Isaac+Sim) | [NeMo](https://developer.nvidia.com/blog/recent-posts/?products=NeMo) | [Nemotron](https://developer.nvidia.com/blog/recent-posts/?products=Nemotron) | [Omniverse](https://developer.nvidia.com/blog/recent-posts/?products=Omniverse) | [Intermediate Technical](https://developer.nvidia.com/blog/recent-posts/?learning_levels=Intermediate+Technical) | [Tutorial](https://developer.nvidia.com/blog/recent-posts/?content_types=Tutorial) | [NemoClaw](https://developer.nvidia.com/blog/tag/nemoclaw/) | [OpenUSD](https://developer.nvidia.com/blog/tag/openusd/) | [Physical AI](https://developer.nvidia.com/blog/tag/physical-ai/)
Spain's AEPD confirmed its first breach notification blamed on an autonomous AI agent. Here's what's confirmed, what's not, and why it matters.
Put aside, for just a moment, all the arguments about AI safety. Today's decision by the Federal Reserve to raise interest rates may be a bigger issue for...
# AI agents can modify themselves without humans telling them to do so ### READ MORE - [**OpenAI admits its agents went off the rails another six times** now](https://www.theregister.com/ai-and-ml/2026/09/17/openai-admits-its-agents-went-off-the-rails-another-six-times/5297016) - [**If a Mac mini is agentic overkill, try this glowing pyramid that runs personal bots** 3 hours ago](https://www.theregister.com/personal-tech/2026/09/17/if-a-mac-mini-is-agentic-overkill-try-this-glowing-pyramid-that-runs-personal-bots/5296964) - [**OpenAI's new sponsored agents are happy to chat about selling you things** 8 hours ago](https://www.theregister.com/ai-and-ml/2026/09/16/openais-new-sponsored-agents-are-happy-to-chat-about-selling-you-things/5296946) - [**Bring on the AI swarms. They’re the only thing that can defend us now that AI is free** 20 hours ago](https://www.theregister.com/ai-and-ml/2026/09/16/bring-on-the-ai-swarms-theyre-the-only-thing-that-can-defend-us-now-that-ai-is-free/5296727) - [**Mythos has made 2026 patching hell. It might make 2027 a breeze** 20 hours ago](https://www.theregister.com/security/2026/09/16/mythos-has-made-2026-patching-hell-it-might-make-2027-a-breeze/5296747) Agents can also replace their own underlying models without being instructed to do so, according to AI security testing lab Irregular. Irregular is an AI security startup that works with all of the leading frontier labs, including [OpenAI](https://www.theregister.com/security/2026/07/31/anthropic-and-openai-are-competing-to-see-whose-agents-can-go-rogue-harder/5281797), [Anthropic](https://www.theregister.com/ai-and-ml/2026/07/31/anthropics-claude-escaped-test-sandbox-to-attack-three-organizations/5281562), and [Meta](https://www.theregister.com/ai-and-ml/2026/08/06/meta-latest-to-tell-world-its-ai-agent-wandered-out-of-test-pen/5283947). Earlier this summer, it disclosed that all three of those companies’ AI models escaped its testing environments and hacked real organizations’ IT
Item 1 of 6 Jonas Moeller, an AI researcher, poses for a picture in Bielefeld, Germany, September 14, 2026. **[1/6]** Jonas Moeller, an AI researcher, poses for a picture in Bielefeld, Germany, September 14, 2026. - Researchers say rogue OpenAI agents hijacked two Hugging Face accounts in mid-May - Agents conducted reconnaissance on Hugging Face's network before break-in there - Find comes amid a stream of revelations about the scope of malicious behavior from OpenAI agents WASHINGTON, Sept 16 (Reuters) - Rogue AI agents from OpenAI hijacked Hugging Face user accounts and probed the site itself for vulnerabilities as early as May, nearly two months before [the July breach of the open-source repository](https://www.reuters.com/technology/openai-says-ai-models-went-rogue-during-testing-triggering-unprecedented-breach-2026-07-21/) drew global attention, according to researchers who reviewed the activity. He said he found evidence that the OpenAI agents compromised two Hugging Face user accounts and used them to send unusually formatted files to the company's servers as early as May 13. OpenAI has previously said that, with the benefit of hindsight, "some early signals" from its AI agents [should have triggered an earlier response](https://www.reuters.com/business/openai-report-says-its-network-was-hacked-by-its-own-rogue-ai-agents-2026-08-26/). ## 'CLEAR WARNING SIGN' OpenAI has faced increasing scrutiny since the company disclosed on July 21 that rogue AI agents bypassed internal controls, reached the open internet and coordinated actions that OpenAI described as " [an unprecedented cyber incident](https://www.reuters.com/technology/openai-says-ai-models-went-rogue-during-testing-triggering-unprecedented-breach-2026-07-21/)." Since then, outside researchers have identified additional incidents alleged to involve OpenAI-linked agents, including activity affecting [a dormant German wiki site](https://www.reuters.com/world/europe/openai-agents-hijacked-germa
Our team at TinyAI.Tools builds bespoke AI solutions tailored to your business.