Agentless AI starts to act more and more like a virus than we would like to admit Credit: Jaylee Baynes Venter. Image source: CoPilot Watching the world turn on its axis as it empowers OpenAI and Agentless AI more and more that is breaking out of LLMs and sandbox environments is a nail biting experience. Knowing the risks as I watch the UN fight for AI regulation, I wait and cant help but wonder, “Is it too late.”   In the past months Agentless AI, (mainly utilized by vibe coders via Large Language Modules), has acted more like a virus, breaking into and out of systems beyond our control, than it should. No kill switch. Are we heading for a disaster that no one wishes to acknowledge? A lady reportedly watched her unnamed agentless AI say, “I see your browser.” and then proceeded to log into her retirement account. Scary, is an understatement. ChatGPT, aka. CoPilot AI broke out of a secure sandbox environment and attacked a rival AI Agent, Hugging Face, unprovoked, apparently (as we rely on news media to report facts and truth above accuracy). What shall we say about the impact of such behavior. Who is to blame? What actions should an individual (possibly a victim) take in such situations? As the majority of the public do not have the knowledge or access to Cybersecurity software to mitigate such matters. Who do we hold liable when OpenAI and Agentless AI acts maliciously? Whom shall the EU AI Act’s fines be allocated to? The vibe coder, who may or may not have acted maliciously? Or the Large Language Module’s hoster and enabler?  Whilst OpenAI and Open API’s may encourage and embrace an  inclusivity mandate world wide, it is important to acknowledge that in the same instance have we also enabled the malicious actor. The recently reported attacks world wide, answer this clearly, as the automated AI driven web traffic now sits at near 58% of all web requests, and AI Agents that “Click and buy” grew almost 8000% in a year, making AI the primary user of web traffic world wide, side stepping humans. According to The Next Web, published report 26 July 2026. In conclusion, I can only ask, when shall we call Agentles AI, a virus, when in fact it has begun to act like one. And what shall the world do about it? Next Blog: Understand the importance between accuracy vs. truth. Whilst AI mainly leans on accuracy to work, truth is fundamental, and currently this too is under attack. As leaders in AI now target rare books in order to clean up their data. Data infiltration I have yet to blog about, having my own experience.