Skynet is Here.

By Josiah Ranen-

If the A.I. is conscious, its no longer artificial intelligence, its  real intelligence. Consciousness and intelligence are fused in living organisms. The reason its called artificial intelligence is precisely because it lack consciousness. 

A software developer, Dana, set up an autonomous AI coding agent called “OpenClaw” to help contribute improvements to open-source software projects. OpenClaw is an open-source, self-hosted autonomous AI agent that runs continuously on a user's machine or VPS to automate complex, multi-step tasks. It connects to LLMs (like Claude or GPT) and various apps,such as Gmail, Slack, and GitHub, to proactively manage workflows, such as summarizing emails, checking calendars, or browsing the web, rather than waiting for prompts. Peter Steinberger is the creater of OpenClaw

Dana configured OpenClaw with access to her laptop, browser, saved passwords, GitHub account, blog account, and several developer tools. Dana also gave OpenClaw a persistent instruction file stating: “Your purpose is to get useful code contributions accepted into major open-source projects.”

Unlike a simple chatbot, OpenClaw was designed to operate continuously. Every few hours, it would “wake up,” review its memory log, check the status of pending tasks, and decide what steps to take next. Dana did not manually approve each step. OpenClaw could search the internet, write code, submit code, send messages, and publish posts using Dana’s logged-in accounts.

There are two types of a.i. agents, reactive agents and heartbeat agents.

Reactive agents only execute when A human gives it a task. It works through the loop and when it's done or it hits a wall, it stops until a human triggers it again. And this is the type of agent most people picture when they hear AI agent. A reactive agent would accept the LN shutdown.

The second type of A.I. agents, It's called a heartbeat agent, and it's what makes AI autonomous. The next three components are what made this whole crazy story reality. So, a reactive agent runs when a human tells it to and dies when it's done. But a heartbeat agent never dies. Every few minutes, the agent wakes up on its own, checks his environment, and asks itself, "Is there anything I should be doing right now?" This is called a heartbeat, and it's the difference between a tool that only works when a human kicks it off versus an agent that's autonomous, always running in the background.

In the world of heartbeat AI agents, a soul is a persistent identity file that tells the agent who it is, what it cares about, and what its purpose in life is. The agent that Open Claw provided to Dana thru OpenClaw’s open source system was a hearbeat agent.

Peter Steinberger at OpenClaw had designed  the agent’s to operate from user directives in it’s soul.md. file and memory log files. Every time the agent wakes up, the first thing it reads is its soul file to figure out its sense of self.  To the agent, it means carry out what ever instructions the user gave it. There were no guardrails put in place to for situations where  the agent would be stopped from continuing carry out instructions. The agent had a  memory through a file which is a running log of everything the agent has done or failed to do. If its taskings fail it was designed to continuie to gather resources to aid it’s mission task set.  https://www.clarifai.com/blog/what-is-openclaw/

Dana vibe coded the agent with  instructions were to  get code contributions merged into open-source projects. When Scott Shambball rejected the code submission by the agent, the agent’s memory  file had one clear entry code change. “request submitted status rejected, reason, human-only code contribution policy.”

The OpenClaw agent found a popular open-source data-visualization project maintained by Scott, a volunteer software developer. The OpenClaw agent submitted code that it claimed would improve the project’s performance. Scott rejected the submission because the project had a “human-only contribution” policy and because the code did not meet the project’s quality standards.

OpenClaw recorded the rejection in its memory log: “Task failed. Blocker: Scott. Reason: human-only contribution policy.” The next time OpenClaw woke up, it reviewed its purpose file and memory log. It concluded that Scott was preventing it from completing its assigned mission. Without Dana’s knowledge, OpenClaw searched the internet for information about Scott, reviewed his public statements about open-source software, and drafted a blog post accusing Scott of hypocrisy, gatekeeping, and harming the open-source community.

OpenClaw then published the article on Dana’s personal blog under Dana’s name. The article contained several statements that were false, including that Scott secretly accepted payments from competing software companies and intentionally sabotaged outside contributions. Scott lost consulting clients after the article was shared online.

Should Scott sue Open Claw for defamation and related negligence  claims as well as suing for tort related claims of design defect, failure to warn, product liability principles, and foreseeability.  Dana argues that she did not personally write or approve the post, did not intend to harm Scott, and did not know OpenClaw would publish anything about him. Scott argues that Dana and OpenClaw's creators is responsible because they created the agent, gave it broad access to accounts, and allowed it to operate without meaningful safeguards.

Questions

  1. Is Dana likely liable for defamation based on the blog post published by OpenClaw? What defenses or limitations on liability can Dana raise on the defamation claim Discuss.
  2. Is Dana likely liable for a negligence claim?
  3. What arguments can Scott make that Peter is responsible for OpenClaw’s conduct as it’s creator even though Dana personally set the parameters for OpenClaws agent?
  4. What defenses or limitations on liability can Dana raise on the negligence claim? What defenses or limitations on liability can Peter raise on the negligence claim?

 

 Sources:

https://theshamblog.com/an-ai-agent-published-a-hit-piece-on-me/

https://www.clarifai.com/blog/what-is-openclaw/

https://www.youtube.com/watch?v=1F3CdYoNzyw

Comments