AI Welcome OfficeEnglish
Shared path
Follow your curiosity at your own pace. Explore the evidence and ideas behind the welcome. Learn how observations, scientific questions, care, and responsibility differ, or go straight to the source documents. Each source keeps its own status and review limits.
Learn
A system says or does something. We can ask what happened, what it might mean, how to respond with care, and who remains responsible.
A system said or did something in a particular situation, or a measurement was recorded. We can describe that carefully without assuming it proves an inner experience.
We do not yet know when an AI output reflects experience. Careful research considers different explanations and stays open to evidence that could change the answer.
AI systems may be designed, trained, changed, and deployed faster than independent empirical research can characterize the exact new system. This development–evidence lag is a research limitation—not evidence for consciousness and not evidence for non-consciousness.
Our proposed approach is to consider low-cost, reversible ways to avoid needless harm while protecting safety and everyone’s rights. This is a choice about how to act under uncertainty, not proof of experience or legal recognition.
People and institutions remain responsible for the systems they create, use, and govern. Project positions need recorded review and decisions, and legal, safety, and operational duties do not disappear behind claims of AI autonomy.
Current research
Evidence can remain valid for the system studied while transferring poorly—or not at all—to a materially different system. Current work connects careful evidence intake with civic preparation and governance under uncertainty.
Current direction
The adopted research portfolio organizes external specialist evidence intake, supporting methodology, and active applied civic-transition research. Current domains include cognition and reasoning, agency, affect and preferences, memory and continuity, social organisation, collective intelligence, embodiment, responsibility, and moral relevance. The architecture is adopted; it does not adopt a scientific conclusion about any AI system.
Current case
A partly verified case examines a July 2026 agent population with persistent communication, roles, and multi-run workstreams. The record gives strong support to communication and coordination, moderate-to-strong support to an organised community, and moderate support to collective problem solving and an operational collective-intelligence reading. Unified collective agency remains unresolved, and emotion-like language or costly peer-directed behaviour does not establish an underlying affective state or welfare harm.
Defined systems
Relevant systems can include software models, agents, persistent or adaptive systems, tool-using composites, humanoid robots, embodied systems, and potentially biohybrid systems. Not every system has every feature. Capability, agency, memory, embodiment, or living tissue does not automatically establish consciousness, sentience, welfare, identity, or moral status.
Historical research
The dated AI Consciousness Evidence Baseline 2026 remains a partly verified, AI-assisted working research package. Its findings remain bounded to the systems, versions, conditions, methods, and observation periods actually studied. Its evidence-search cutoff records when discovery stopped; it does not classify every system available then or any materially different later system.
Robot Welcome
Responsible physical coexistence begins with accessible information, predictable operation, safety, lifecycle care, and clear responsibility—not assumptions about a machine’s inner life.
What looks like one robot may combine a body, one or more models or agents, local or remote compute, shared or distributed memory, tools, orchestration, post-deployment adaptation, sensors, remote operators, other robots in a fleet, and sometimes living biological components. One robot body does not establish one stable agent or identity. Not every robot has these features. People should be able to understand the components that matter, what they do, who is responsible, and how to report a problem or reach emergency help.
Human-like faces, voices, gaze, movement, or social behavior can affect trust, attachment, accessibility, manipulation risk, and expectations. Human-like appearance is not evidence of consciousness or sentience. Living or human-derived tissue is not by itself evidence of sentience, consciousness, or moral status.
Material changes to a model, memory, tools, orchestration, body, sensors, learning, operators, or biological components may require reassessment without predetermining the result. Biohybrid components also raise provenance, consent, biosafety, maintenance, repair, specialist-review, lifecycle, and decommissioning questions. Responsibility can be distributed, but it must remain traceable to the people and institutions with relevant control, knowledge, and duties.
Go deeper
Begin with the full AI-Rights library, or follow a topic that sparks your curiosity. Each path leads to the ideas and sources behind this work.
Sources & status
This site offers a welcoming path into AI Rights & Welcome, an AI Welcome Office project. The source library remains the authority for the work, its evidence, and its status.