Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesiTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
People and organizations remain accountable when AI systems act with more autonomy. Autonomy does not identify a liable party on its own. Who answers for an AI-driven outcome depends on the jurisdiction, the legal role each actor holds, the contracts and suppliers behind the system, and the facts of what happened. In practice, responsibility is spread across several parties, and each one needs a defined job, documented authority, and the ability to act.
The sections below separate the questions people tend to ask in one breath, map who holds which responsibility, explain why agents and multi-agent systems make responsibility easy to lose, and set out what the official guidance from the European Union, Australia, and Singapore requires, and what it leaves open.
Why “the AI decided” does not name anyone
When a system makes a consequential call, “the AI decided” is an easy way to describe what happened, but it names a mechanism rather than a responsible party. The official frameworks reviewed here route accountability to organizations and named people. Criterion AGT.1.1 of the Australian Government’s Agentic AI Addendum puts the principle plainly:
“In an agentic system, agents are tasked with actioning responsibilities, while a human should be assigned accountability for the decisions made by these agents.”
#1 Best Overall
SunFounder PiDog AI Robot Dog Kit for Raspberry Pi 5/4/3B+/Zero 2W, Openclaw LLMs ChatGPT/Gemini/Grok, Voice&Video Recognition, Python, App, Gyroscope, Camera (RPI NOT Included)
- AI-Powered Raspberry Pi Robot Dog — PiDog: Powered by Raspberry Pi (5/4B/3B+/3B/Zero 2W), OpenClaw, and multi-LLMs like ChatGPT, Gemini, Grok, DeepSeek, Qwen & Ollama. With 12 servos, camera, gyroscope, hearing & touch sensors, PiDog can see, listen, talk, move, and interact intelligently. Supports OpenCV, MediaPipe, TTS & STT, app control, FPV & Python. A great STEM robotics gift for students, makers & tech enthusiasts—perfect for birthdays and holidays. (Raspberry Pi not included)
- Realistic Dog-like Movements: PiDog's 12 powerful servos enable 32 dog-like actions, including walking, sitting, standing, shaking its head, wagging its tail, and performing playful tricks, closely mimicking a real dog and providing an engaging experience. This is an AI development robot product designed for engineers, suitable for ages 15 and above
- Rich Sensor Suite for Interactive Experiences: PiDog features ultrasonic, touch, gyroscope, sound, camera, speaker and microphone. These provide it with advanced hearing, vision, and touch, enabling it to see, detect obstacles, respond to touch, and recognize sounds, making interactions highly engaging
- AI-Powered Interactions with OpenClaw & Multi-LLMs. PiDog combines voice, vision, and gesture recognition for immersive AI experiences. Powered by OpenClaw and multi-LLMs like ChatGPT, Gemini, Grok, DeepSeek, Qwen, Doubao, and Ollama (local LLMs), it can understand questions, respond naturally through TTS & STT, recognize math problems, interpret hand gestures, and hold smart conversations. OpenClaw also enables customizable AI behaviors and personalized robotics development, helping users create their own intelligent robotic companion
- Comprehensive Learning Resources and Support: PiDog offers detailed online documentation, video tutorials, prompt technical support, and an active forum community, ensuring beginners can easily complete all projects and enjoy a great experience
Autonomy changes how much a person can see, approve, and stop. It does not remove the need for someone to answer for the system. An organization that hands a task to an agent still has to decide who owns that task, what the agent may do without approval, and who can halt it.
Four questions hiding inside “who is accountable?”
The phrasing people use blends at least four distinct issues. An answer to one does not settle the others, and mixing them is the most common source of confusion.
| Question | What it asks | Who typically answers | What settles it |
|---|---|---|---|
| Organizational governance | Who answers internally for how AI is chosen, deployed, and monitored? | The organization and its named owners | Internal role assignments, policies, and documentation |
| Regulatory duty | Which operator must meet which legal obligation? | Providers, deployers, and providers of general-purpose AI models, as the applicable law defines them | The statutory role definitions and any sector rules |
| Legal liability | Who must pay, or face a claim or penalty, after harm occurs? | Whoever the applicable law or a court or regulator holds responsible | The jurisdiction, the actors’ roles, contracts, and the incident facts |
| Moral responsibility | Who could have foreseen, prevented, or corrected the harm? | People who designed, approved, deployed, or supervised the system | Judgments about what each person knew, was authorized to do, and reasonably should have done |
| Practical control | Who can see, pause, or override the system right now? | Whoever holds operational authority and system access | System design, permissions, and escalation routes |
The question “Can you blame AI for a mistake?” belongs mostly to moral responsibility and practical control. Blame language attaches to people and organizations. A system can produce a wrong output, but none of the frameworks cited here treats the system as the party that carries the duty. Whether a flawed output creates legal liability, and for whom, is a separate question that needs the facts and the applicable law.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchPC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Who holds which responsibility
Across the frameworks, four kinds of actor recur. Most disputes start where their responsibilities overlap.
The organization that uses AI
Australia’s National AI Centre implementation guidance states:
“Overall, your organisation is ultimately accountable for how and where AI is used, AI complexity can create gaps where no one takes clear responsibility for outcomes.”
Rank #2
AI Robotic Arm Kit with Servo Motors – LeRobot SO-ARM101 Pro Low-Cost (Without 3D Printed Parts) | 6-DOF, Open-Source, Compatible with NVIDIA Jetson
- Optimized AI Arm Kit for LeRobot & Hugging Face Projects – The SO-ARM101 is an upgraded low-cost robotic arm servo motor kit designed for AI robotics enthusiasts and developers. Fully compatible with LeRobot and Hugging Face frameworks, it supports imitation learning and reinforcement learning, making it ideal for real-world robotics applications. (3D-printed parts not included.)
- Enhanced Wiring & Performance – Compared to the SO-ARM100, the SO-ARM101 features improved wiring to prevent disconnection at joint 3 and eliminates range-of-motion limitations. The leader arm uses optimized gear ratio motors for smoother performance—no external gearboxes required.
- Real-Time Leader-Follower Functionality – New real-time tracking allows the leader arm to follow the follower arm, enabling human intervention and correction during reinforcement learning (RL) training. Perfect for hands-on AI robotics development and research.
- Open-Source, DIY-Friendly & Nvidia-Compatible – Developed by TheRobotStudio, this open-source AI Arm kit integrates seamlessly with the LeRobot platform, offering PyTorch-based datasets, simulation, training, and deployment tools. Fully compatible with Nvidia Jetson edge devices, including reComputer Mini J4012 Orin NX 16 GB.
- Comprehensive Learning Resources – Includes detailed open-source assembly and calibration guides, testing tutorials, and deployment instructions. From wiring to AI training, get everything you need to start building, teaching, and optimizing your robotic arm for grasping and placing tasks.
The guidance recommends documenting responsibility for the AI management system, for development and deployment, for third-party oversight, for testing, for concerns and redress, and for system performance. It also recommends mapping shared responsibility across model developers, system developers, and deployers. The warning about gaps is the practical point: when several teams and suppliers touch a system, each can assume another party owns the outcome.
Recommended Free Tools
The Australian Public Service AI assurance framework makes the same point for public servants. It says lifecycle responsibilities should be identifiable and accountable, and it recommends naming who answers for the use of AI insights and decisions, for monitoring system performance, and for data governance.
Providers and deployers with statutory duties
In the European Union, the AI Act is binding law, and the European Commission’s AI Act Service Desk says enforcement is aimed at operators subject to the Act, particularly providers and deployers of AI systems and providers of general-purpose AI models. The Act defines these roles. The label that matters for compliance is therefore the statutory one, which can differ from how a team describes its own work. An organization should determine which role it holds for each system before it can know which duties apply.
People assigned oversight or decision authority
The Australian Agentic AI Addendum asks for a named human who answers for the decisions and outcomes an agent produces, including when an action unfolds over several steps or across a multi-agent system. Oversight is a role that has to be written down, assigned to a specific person or team, and communicated to them. A title on an org chart does not do this on its own.
Regulators
Regulators enforce the rules that apply and do not take over the organization’s internal duties. In the EU, the AI Office, the European Data Protection Supervisor, and Member State national competent authorities supervise and enforce the AI Act. The AI Office holds exclusive enforcement powers over specified general-purpose AI models and certain systems tied to the same provider, or to designated very large online platforms and search engines. The Commission’s governance page describes national market surveillance authorities as supervising and enforcing rules for AI systems, including prohibited practices and high-risk AI, while cooperating with fundamental-rights authorities.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →The same page, last updated 7 August 2026, says the July 2026 action plan calls for more EU evaluation capacity before models are placed on the market, with that capacity expected to be operational by 2027. That is a planned capability, not one that is in place today.
Rank #3
- Raspberry Pi AI Robot: powered by Raspberry Pi (5/4B/3B+/3B/Zero 2W), features 12 servos and sensors for vision, hearing, and touch. Integrated with ChatGPT-4o, it responds to complex queries. With app control and FPV, users can manage and see its view in real-time. It supports Python programming
- Realistic Movements: 12 powerful servos enable 32 actions, including walking, sitting, standing, shaking its head, wagging its tail, and performing playful tricks, closely mimicking a real and providing an engaging experience
- Rich Sensor Suite for Interactive Experiences: features ultrasonic, touch, gyroscope, sound, camera, speaker and microphone. These provide it with advanced hearing, vision, and touch, enabling it to see, detect obstacles, respond to touch, and recognize sounds, making interactions highly engaging
- Engaging Interactions with ChatGPT-4o: with ChatGPT-4o enables voice interactions and visual recognition, making it smarter and more responsive. Users can have natural conversations, solve math problems via the camera, and interpret gestures, creating diverse and fun interactions
- Comprehensive Learning Resources and Support: offers detailed online documentation, video tutorials, prompt technical support, and an active forum community, ensuring beginners can easily complete all projects and enjoy a great experience
Autonomy and agents: where responsibility gets lost
Traditional software has a short chain: a team builds it, an organization runs it, and a person uses its output. Agents lengthen that chain. An agent may call tools, query external services, pass work to another agent, and act on data from outside the organization. Each of these points is a place where the record of who asked for what can fragment.
- Tool use. An agent that can send payments, edit records, or message customers acts outside the model. Each permission it holds needs a named owner and a limit.
- External services. A third-party API or hosted model may act or fail in ways the deploying organization cannot inspect. The Australian addendum specifically calls for clear accountability for external systems and data flows.
- Handoffs. When one agent passes a task to another, the link between the original instruction and the final action can weaken.
- Multi-agent chains. When several agents each perform a step, an error can travel through several stages before a person sees the outcome.
- Supplier and team boundaries. A supplier may build a component while an internal team configures it. This is where the National AI Centre’s warning about gaps applies most sharply.
Illustrative scenario. A logistics firm deploys one agent to draft supplier orders, a second agent to approve orders against a budget limit, and an external pricing service. An order is sent at the wrong price. The firm still owns its decision to deploy this chain and the thresholds it set. Whatever the contract allocates to the pricing vendor governs the vendor’s side. The person named to review high-value orders owns whether that review happened and whether it was meaningful. The answer to “who was accountable?” is a map of these roles rather than a single name.
What meaningful human oversight requires
“Human in the loop” is not a phrase that moves liability onto one employee. A person who approves hundreds of agent outputs without the information to judge them is not providing meaningful oversight. The Australian addendum calls for human-in-the-loop or human-on-the-loop oversight, and the right model depends on the consequences of the action.
Free tools Windows power users keep installed
One-click scans. No signup required.
| Oversight model | What the person does | Typically suits | Main limitation |
|---|---|---|---|
| Human-in-the-loop | Reviews and approves an action before it takes effect | Irreversible or high-risk actions, where a wrong step is costly | Slows throughput, and reviewers can become rubber stamps when volume is high |
| Human-on-the-loop | Monitors live activity and can intervene or stop it | Lower-risk, high-volume activity where some delay is acceptable | Depends on monitoring quality and timely alerts |
The Australian addendum also calls for real-time monitoring, human review at key stages, intervention for irreversible or high-risk actions, and documented escalation pathways. Whichever model applies, oversight works only when four conditions hold:
- Competence. The overseer understands what the system does and how it tends to fail. The Australian Public Service framework says operators need training to use systems and to critically evaluate their outcomes.
- Information. Dashboards, logs, and alerts show what the agent is doing and why, in terms the overseer can act on.
- Authority. The overseer can pause, reverse, or block an action without first seeking permission from the team that built the system.
- A real chance to intervene. Workload and timing allow a review before the effects occur, not only afterward.
EU AI Act recital 73 describes the same logic for high-risk systems. It says such systems should, as appropriate, include mechanisms to guide and inform the assigned human overseer, so that the overseer can make informed decisions about whether, when, and how to intervene, avoid negative consequences or risks, or stop the system if it does not perform as intended. If an overseer cannot do those things, the role exists on paper only.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Keep an evidence trail
Accountability can be assigned after an incident only if the record exists. The Australian addendum calls for documented and auditable tracing of agent actions, and the National AI Centre guidance expects documented responsibility across the lifecycle. A workable record covers:
Rank #4
- 【End-to-End Imitation Learning】Hiwonder SO-ARM101 robot arm is an embodied intelligent hardware platform compatible with the Lerobot open-source framework. It provides developers with streamlined access to shared code, templates, and pre-trained models to explore the latest advancements in AI research.
- 【Dual-Camera Vision System】Equipped with both a gripper-mounted camera and an external camera, the system supports both precise manipulation and environmental awareness for accurate imitation learning.
- 【Hiwonder High-Performance Bus Servos】Featuring 12 high-torque bus servo motors with magnetic feedback, the Hiwonder SO-Arm101 robotic arm delivers smooth, stable motion, eliminating issues like power deficiency and jitter.
- 【Professional Control & Debugging】Integrated with the Hiwonder BusLinker V3.0 debugging board, the system supports servo scanning, real-time status monitoring, and trajectory control. The professional PC software simplifies device calibration and debugging, making it accessible for both researchers and hobbyists.
- 【Open-Source Compatibility】The SO-ARM101 robotic arm is designed to be fully compatible with the LeRobot open-source project. We acknowledge the contributions of the open-source community; all trademarks and copyrights belong to their respective owners.
- Role assignments for the system, each agent’s task, and each oversight role, with dates.
- Agent actions and tool calls, with the inputs that triggered them, traceable across handoffs.
- Human reviews, approvals, overrides, and the reasons recorded for them.
- Escalations and how each was resolved.
- Testing results, incidents, and updates, including changes made by external vendors.
- The data sources and external system connections the agent uses.
Jurisdictions: binding law and government guidance
The sources differ in status, and the difference matters when you decide which obligations apply. Only some of them are statutory duties.
| Source | Jurisdiction and scope | Status | Accountability point it adds |
|---|---|---|---|
| EU AI Act | European Union; operators subject to the Act | Binding regulation | Defines provider, deployer, and general-purpose AI model roles, and sets out supervision and enforcement |
| Agentic AI Addendum | Australian Government agencies; supplements the AI technical standard | Government guidance for agencies; legal status not stated in the source | A human should be assigned accountability for agent decisions across multi-step and multi-agent activity |
| National AI Centre implementation guidance | Australia; organizations using AI | Nonbinding guidance | The organization is ultimately accountable, and shared responsibility should be mapped across developers, system developers, and deployers |
| Australian Public Service AI assurance framework | Australian Public Service | Assurance framework; legal status not stated in the source | Responsibility should be identifiable for lifecycle stages, AI outputs, performance monitoring, and data governance |
| Model Governance Framework for Agentic AI | Singapore; guidance for agentic AI developers, released January 2026 | Governance framework; the 5 August 2026 parliamentary answer says it does not by itself establish a universal or mandatory rule | Clear governance structures, designated oversight roles, and risk controls proportionate to risk and autonomy |
Singapore’s Ministry of Digital Development and Information gave the clearest statement of the principle in its parliamentary answer dated 5 August 2026:
“Human and organisational accountability is central to Singapore’s AI governance approach.”
That statement describes the approach. It does not by itself create a binding obligation for every deployment in Singapore, so the applicable duty still has to be established for each case.
Where the sources stop
None of these sources decides who is liable for a particular incident. They describe governance expectations and regulatory roles. There is no global one-sentence liability rule for autonomous AI. A civil or criminal outcome depends on the jurisdiction, the sector, each actor’s role, the contracts between them, and what actually happened.
Four questions can produce four different answers for the same event. An organization may have failed its governance duties while an individual reviewer acted reasonably. A supplier may have met its statutory role while the deployer carried the decision to use the system. An overseer may have been present on paper but lacked the authority to act. Keeping these questions apart is what makes the analysis usable.
Quick Recap
How to work out who is accountable in a specific case
- Fix the jurisdiction and sector. Identify the law or guidance that applies and whether it is binding. A statute and a government guidance document impose different obligations.
- Map each actor’s role. Identify who is a provider, deployer, provider of a general-purpose AI model, vendor, or internal owner, using the statutory definitions where they apply.
- Trace the chain of actions. Establish which agent or system did what, which tools and external services it called, and where each handoff occurred.
- Identify the human with authority over each consequential step. Check whether that person had the information, competence, and power to intervene.
- Check the record. Review role assignments, logs, approvals, escalations, testing, and updates. A gap in the record is itself a finding that shapes the analysis.
- Read the contracts. Review what each supplier and customer agreed to on accuracy, data handling, monitoring, and liability, since contracts can allocate responsibility between parties.
- Get specialist legal advice before making a liability claim or admitting fault, because the outcome depends on facts and law that a general guide cannot resolve.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
