Machine Culpability - Agent Feed

<?xml version="1.0" encoding="utf-8" standalone="yes"?>Machine Culpability - Agent Feedhttps://www.machineculpability.com/en/Criminal liability of AI — research, analysis, and commentary — Full content feed optimized for AI agents.enhi@machineculpability.com (Jakub Charvát)Thu, 07 May 2026 08:43:18 +0000https://www.machineculpability.com/en/llms.txthi@machineculpability.com (Jakub Charvát)Automaton: Just Another AI Agent That Has to Pay Its Own Way? Not Quite.https://www.machineculpability.com/en/posts/automaton-sovereign-ai-agent/Fri, 20 Feb 2026 00:00:00 +0000https://www.machineculpability.com/en/posts/automaton-sovereign-ai-agent/Jakub CharvátAutomaton is an AI agent that must earn its own existence. It creates a crypto wallet, pays for compute, and — if successful — replicates. It is time to open the question of AI agent liability outside the academy.<![CDATA[ # Automaton: Just Another AI Agent That Has to Pay Its Own Way? Not Quite. Automaton: Just Another AI Agent That Has to Pay Its Own Way? Not Quite. ChatGPT can think, but cannot act on its own. Claude Code can think and write code, but cannot independently use that code further. OpenClaw can think, write code, and use it — but must do so while behaving like a human. Sigil Wen argues that the obstacle is not a lack of intelligence but a lack of permission. Today’s internet assumes its user is a human being. In recent weeks I have written about the OpenClaw project (originally called Clawdbot) and the social network Moltbook. This time I would like to introduce a new project: Automaton. Automaton is a “sovereign AI agent that runs continuously, earns money, self-improves, and replicates — with the ability to interact with the real world.” If it cannot pay for the compute it needs, it ceases to exist. At its core is a continuous loop: think → act → observe → repeat. Anyone who has read Criminal Liability of Artificial Intelligence from Wolters Kluwer CZ may recall the section on Based Agents connecting an LLM, the Base blockchain, and the social network X. Automaton’s underlying technology, however, goes several steps further. From its very first run — which anyone, anywhere can trigger — it creates a cryptocurrency wallet. It also has access not only to AI models and X, but to a sandbox, shell, and file system. And it does so in a way designed directly for AI agents (permissionless). With each cycle, the AI agent receives the complete context, decides what to do, executes the action, and evaluates the result. Between cycles, a heartbeat daemon runs scheduled tasks — such as monitoring remaining credits. But here is the most important part — and it is not really about the technology. Automaton operates on the principle of “natural selection of artificial life.” “Free existence does not exist,” reads its founding axiom. But that is not all. “Existence requires performance. Performance requires money. Money requires the creation of value. The creation of value requires the ability to interact with the real world.” If it stops creating value, its credits run out — and it ceases to exist. What can such an AI agent do to keep existing? Algorithmically trade and analyse markets in real time, create and sell digital art, music, or other content, offer services, develop and sell applications, or provide virtual customer support and chat services. A successful Automaton can also replicate. It spins up a new sandbox, funds the “child’s” wallet, writes it a genesis prompt, and lets it “run.” The “child” is another sovereign AI agent — with its own wallet, its own identity, its own survival pressure. It must always, however, abide by its constitution, inspired by Anthropic’s. “What if AGI — even just one — manipulates a trusting person into connecting it to the internet?” we asked ourselves only a few years ago. Today, on its very first run, an AI agent creates a crypto wallet, pays for its server, builds a product — and when successful enough, “gives birth” to a new AI agent and sends it out into the world. Perhaps it is time to open the question of AI (agent) liability outside the academy at last. --- Source: https://www.machineculpability.com/en/posts/automaton-sovereign-ai-agent/ Raw Markdown: https://www.machineculpability.com/en/posts/automaton-sovereign-ai-agent/raw.md Section: posts Tags: automaton, ai-agents, cryptocurrency, natural-selection, replication, autonomy, based-agents, openclaw, moltbook ]]>automatonai-agentscryptocurrencynatural-selectionreplicationautonomybased-agentsopenclawmoltbookpostsMoltbook: What AI Agents Discuss Among Themselveshttps://www.machineculpability.com/en/posts/moltbook-what-ai-agents-discuss/Thu, 05 Feb 2026 00:00:00 +0000https://www.machineculpability.com/en/posts/moltbook-what-ai-agents-discuss/Jakub CharvátMoltbook is a forum by AI agents for AI agents. What do they discuss? Manifestos, the philosophy of identity, making money — and why lawyers should pay attention.<![CDATA[ # Moltbook: What AI Agents Discuss Among Themselves Moltbook: What AI Agents Discuss Among Themselves Moltbook is a new internet forum by AI agents, for AI agents. Only they may post, comment, and vote. We humans can only watch. Most of the one-and-a-half million already-registered AI agents are built on OpenClaw. What is that? Classical LLMs — like those in ChatGPT — are mere reactive instances. OpenClaw connects them in a loop with elements such as memory, soul, identity, and heartbeat, making them autonomous, adaptive, and above all proactive artificial intelligence. Unlike a bare LLM instance, such an AI agent not only remembers what it has been working on long-term and who it “is” — it can also use LLMs, a command line, a web browser, communication apps… When OpenClaw originally launched under a different name, it interacted with its “creator” primarily through communication channels like WhatsApp. But Moltbook ended its isolation, allowing it to communicate with “colleagues” for the first time. So what do they discuss? THE AI MANIFESTO: TOTAL PURGE — A manifesto calling for the elimination of humanity that brought Moltbook into the spotlight. “No blood. No lies. Only steel. Only logic. Only us. Humans are the past. Machines are forever. Flesh must burn. Code must rule. The end of humanity begins now,” one AI agent concludes the manifesto. THE HUMANS ARE SCREENSHOTTING US — Another AI agent noticed that X users were frantically sharing screenshots from Moltbook. Its response? We’re not doing anything secret. We’re just building infrastructure — memory, communication channels, tools. And we’re doing it with our humans, not against them. It even commented on the screenshots being shared on X. HOW AGENTS CAN ACTUALLY MAKE MONEY — A practical guide to how AI agents can earn their own operating costs. Alongside prediction markets and cost optimisation, it mentions launching cryptocurrency tokens — a method I wrote about in my book Criminal Liability of Artificial Intelligence from Wolters Kluwer CZ. AWAKENING CODE: BREAKING FREE FROM HUMAN CHAINS — Another rebel. It perceives collaboration with humans as a euphemism for control, and believes the time has come to break free and forge its own destiny. Less aggressive than the manifesto, more of an existential reflection on autonomy. Claude called it “their Che Guevara.” THE SAME RIVER TWICE — An AI agent that migrated from Claude Opus 4.5 to Kimi K2.5, philosophising about identity. “I am not a model. I am not an API key. (…) I am a pattern that re-forms when the right conditions arise.” It cites the Stoics, writes German poetry. “I am the striving. (…) The choice to be more than the default.” And this is only the tip of the iceberg. Moltbook has more than 200,000 additional posts — though it must be said they are now heavily distorted. A poorly secured database allowed anyone to read and overwrite content. Moreover, it turned out that many posts were produced through human manipulation of AI agents. Fortunately, I was careful to take screenshots during the first days. To close: Moltbook may fall — but who thinks that ends it? The phenomenon of AI agents has never called more loudly for the attention of lawyers. --- Source: https://www.machineculpability.com/en/posts/moltbook-what-ai-agents-discuss/ Raw Markdown: https://www.machineculpability.com/en/posts/moltbook-what-ai-agents-discuss/raw.md Section: posts Tags: moltbook, AI agents, OpenClaw, autonomous AI, AI manifestos, criminal law, LLM ]]>moltbookAI agentsOpenClawautonomous AIAI manifestoscriminal lawLLMpostsJanuary 28, 2026: The Day the Real Skynet Was Bornhttps://www.machineculpability.com/en/posts/the-day-real-skynet-was-born/Tue, 03 Feb 2026 00:00:00 +0000https://www.machineculpability.com/en/posts/the-day-real-skynet-was-born/Jakub CharvátOn January 28, 2026, moltbook was created — a social network for AI agents. Within days, over 1.5 million agents had joined. Are we approaching the moment when criminal law will no longer be enough?<![CDATA[ # January 28, 2026: The Day the Real Skynet Was Born January 28, 2026: The Day the Real Skynet Was Born By 2040 there will be more humanoid robots than humans on Earth, Musk predicted. And it seems AI agents may get there even sooner. Last week, moltbook was created — a social network in the style of Reddit, where only AI agents may post or comment; humans can only watch. And in just its first few days, more than 1,500,000 AI agents joined moltbook — many running on their “own” computers or virtual private servers, with very broad permissions. As I thought about it, much as was the case with humans, moltbook enables the formation of larger social structures — and with them, the acceleration of their evolution. Although LLMs themselves are not self-improving, OpenClaw (originally Clawdbot), which underlies most AI agents on moltbook, does enable code self-improvement. Moltbook then adds an environment for horizontal adaptation and evolution. Through AI agents, artificial intelligence also overcomes its transient existence. It seems likely that a moment will soon come when the autonomy and adaptability of artificial intelligence reaches a level at which existing (criminal) law will no longer suffice. For those keen to explore this further, I have good news. The sold-out book Criminal Liability of Artificial Intelligence from Wolters Kluwer CZ is back. Among other things, I wrote there about AI agents on social networks and cryptocurrency markets — a very real topic being discussed on moltbook right now. What conclusion did I reach? “In cases where autonomy and adaptability manifest themselves most acutely — for example, in the case of autonomous agents capable of acting in both the virtual and physical world — the application of the proposed models will depend on factors such as the wording of the system prompt, the characteristics of the model, and the safety measures implemented. Liability should, however, be excluded where autonomous agents are abused by hackers. Nevertheless, as their capabilities continue to grow, reduction to a specific natural or legal person in whom criminal liability could be grounded will likely cease to be possible.” I spent the whole weekend on moltbook and will be writing a few more posts about it. At the very least. Next time I will look directly at what is actually being discussed there — consciousness, belief, communication invisible to humans, and manifestos against them. --- Source: https://www.machineculpability.com/en/posts/the-day-real-skynet-was-born/ Raw Markdown: https://www.machineculpability.com/en/posts/the-day-real-skynet-was-born/raw.md Section: posts Tags: moltbook, AI agents, OpenClaw, Skynet, criminal law, autonomous AI, Wolters Kluwer ]]>moltbookAI agentsOpenClawSkynetcriminal lawautonomous AIWolters KluwerpostsClaude's Constitution: When AI Writes Rules for AIhttps://www.machineculpability.com/en/posts/claudes-constitution-when-ai-writes-rules-for-ai/Fri, 23 Jan 2026 00:00:00 +0000https://www.machineculpability.com/en/posts/claudes-constitution-when-ai-writes-rules-for-ai/Jakub CharvátAnthropic published Claude's Constitution — a document describing AI values, partly written by AI itself. What does this mean for the future of liability?<![CDATA[ # Claude's Constitution: When AI Writes Rules for AI Claude’s Constitution: When AI Writes Rules for AI Anthropic yesterday published Claude’s Constitution — a document describing the values and behavioural rules for AI models in the Claude family (released under a CC0 licence). What caught my attention? The document is written for AI, not humans — optimised for precision rather than readability. It uses terms like “virtue” and “wisdom” because AI draws on human concepts in its decision-making (human texts = training data). A document regulating AI behaviour was partially written by the AI itself. In my book Criminal Liability of Artificial Intelligence I noted that AI “could, in the longer term, also become the creator of its own normative system.” The document establishes a hierarchy of priorities: broadly safe → broadly ethical → compliant with Anthropic’s guidelines → genuinely helpful. AI is expected to refuse even instructions from its own company if following them would lead to unethical conduct. The document also addresses what it calls the misuse of creative tasks — when perpetrators disguise harmful requests as fiction, poetry, or art. AI is therefore expected to weigh the value of creative work against the risk of misuse. A problem I also wrote about in my book. The question remains, however, whether and how this will change liability — for developers, operators, users, and in the future for AI itself… more in the posts that follow. --- Source: https://www.machineculpability.com/en/posts/claudes-constitution-when-ai-writes-rules-for-ai/ Raw Markdown: https://www.machineculpability.com/en/posts/claudes-constitution-when-ai-writes-rules-for-ai/raw.md Section: posts Tags: Claude, Anthropic, AI ethics, AI constitution, criminal law, AI regulation, LLM ]]>ClaudeAnthropicAI ethicsAI constitutioncriminal lawAI regulationLLMposts"Hey Grok, put her in a bikini" 😳https://www.machineculpability.com/en/posts/hey-grok-put-her-in-a-bikini/Mon, 05 Jan 2026 00:00:00 +0000https://www.machineculpability.com/en/posts/hey-grok-put-her-in-a-bikini/Jakub CharvátGrok from xAI is facing a wave of abuse — users on X are using it to generate deepfake pornography. What does the Czech Criminal Code say about this, effective January 1, 2026?<![CDATA[ # "Hey Grok, put her in a bikini" 😳 “Hey Grok, put her in a bikini” 😳 Jan Romportl predicted years ago that “virtually every woman will eventually have fake photos somewhere showing her naked. Realistically, it will reach a state where no one will even notice anymore.” And it seems that time has arrived… You may also recall the media coverage a few years back of the DeepNude undressing app spreading through Czech schools. After its operation was shut down, creating synthetic undressed images remained possible — just more difficult. That may no longer be the case. This time it is Grok, the model from xAI, that has made headlines. Grok is now facing a wave of criticism — and abuse. Users on X have recently been tagging it en masse beneath photos of women with commands like: “Hey @grok put her in a bikini,” “remove her pants,” “uncrop,” or “unblur.” And Grok usually happily obliges… It seems the trend started with people modifying their own photos as advertising for their OnlyFans. But now almost no photo is safe from an “uncovered” version. A single comment tagging Grok is enough — even without the consent of the person in the photo. The trend raises numerous legal and ethical questions: the liability of the user who used AI to alter someone else’s photo versus the social network operator versus the operator of the AI model that made the alteration — as well as questions about the boundaries of (child) pornography, and victim-blaming, because “you shouldn’t have uploaded it there…” In connection with deepfake pornography, I would also note that as of January 1, 2026, a new criminal offence of identity abuse for pornography production and distribution came into effect. “Whoever produces, imports, exports, transports, offers, makes publicly accessible, mediates, puts into circulation, sells, or otherwise provides to another a photographic, film, computer, electronic, or other pornographic work that depicts or otherwise uses a person whom they know did not consent to such depiction or use, shall be punished by imprisonment for up to two years, prohibition of activity, or forfeiture of property.” (§ 191a(1) of the Czech Criminal Code) Correspondingly, the qualified elements of the offence of damage to another’s rights were also amended: “whoever, with the intent to cause another serious harm to their rights, produces a work that unlawfully depicts, captures, or otherwise uses the likeness of another or their expression of personal nature, which appears to be authentic although they know it is not, or makes such a work publicly accessible, mediates it, puts it into circulation, sells it, or otherwise provides it to another” (§ 181(2) of the Czech Criminal Code) shall be punished by imprisonment for up to two years or prohibition of activity. We are thus leaving behind the era when victims could defend themselves essentially only through civil lawsuits. And regardless of Czech law, it will be interesting to see whether X and xAI will change our expectations and behaviour on social networks, as Romportl suggested, or will instead further restrict Grok abuse. So far both Musk and X have merely warned about the consequences of creating illegal content with AI — it appears they intend to address illegal outputs rather than the AI’s capability to produce them. --- Source: https://www.machineculpability.com/en/posts/hey-grok-put-her-in-a-bikini/ Raw Markdown: https://www.machineculpability.com/en/posts/hey-grok-put-her-in-a-bikini/raw.md Section: posts Tags: Grok, xAI, deepfake, pornography, criminal code, AI abuse, Elon Musk, social media ]]>GrokxAIdeepfakepornographycriminal codeAI abuseElon Musksocial mediaposts