The first ethical standard for AI companions — free to read
IEEE 7014.1 is the world’s first international standard for the ethical design of empathic, partner-based AI — the technology behind AI companions, personal AIs and assistants. Built by around 70 experts across many countries, it sets out 29 practical recommendations, and it’s free to read.
Why this matters now
AI that performs empathy is scaling faster than our rules for it. Millions of people now talk to AI companions, partners and assistants — for company, support, coaching, even grief. The benefits are real, and so are the risks: dependency, manipulation, deception, and harm to children and people who are vulnerable.
Lawmakers are racing to respond. In the first weeks of 2026 alone, dozens of chatbot bills were introduced across US states; California’s companion-chatbot law is in force; New York has moved to bar companion bots for minors; the EU AI Act’s rules for general-purpose AI take effect from August 2026; and China has rules on anthropomorphism. But across this fast-moving patchwork, one question keeps recurring: what does good practice actually look like?
What IEEE 7014.1 is
IEEE 7014.1 is an internationally agreed Recommended Practice for the ethical use of emulated empathy in general-purpose AI systems built for human–AI partnership — the technology marketed as companions, personal AIs, co-pilots, agents and assistants.
It was developed over 18 months through IEEE’s consensus process by a working group of around 70 people from many countries and disciplines, and it extends IEEE 7014, the parent standard on emulated empathy. It doesn’t stop at principles: it gives 29 concrete recommendations that apply right across a product’s lifecycle, from first concept to decommissioning.
Weak empathy, strong empathy — the idea at its heart
The standard introduces a simple distinction that helps everyone — designers, regulators and the public — talk clearly about these systems:
Weak empathy is what machines can do: sense, read and respond to human emotion, and interact convincingly.
Strong empathy is what people have: felt experience, solidarity and a genuine sense of care. An AI can convincingly perform empathy without ever experiencing it.
That difference — performing care versus feeling it — is exactly what designers must handle responsibly, and what users deserve to understand.
What the standard covers
Its 29 recommended practices address the issues at the core of the AI companions debate, including:
Deception, and telling people when they’re talking to a machine
Influence, manipulation and nudging
Children, and protecting young and vulnerable users
Emotional dependency and unhealthy relationships
Intimate and sexual interaction
Sycophancy and engagement-driven flattery
‘Ghostbots’ — digital resurrection of the deceased — grief and consent
Mental-health solutionism and crisis situations
Privacy, fiduciary responsibility, bias and cultural fairness
Who it’s for
For regulators, policymakers and courts. A consensus, internationally balloted benchmark of good practice you can cite — evidence of what reasonable design looks like, filling the gap between high-level law and concrete practice.
For developers and AI companies. A defensible standard to build and audit against, that maps onto the obligations you already face and helps reduce legal and reputational risk across a fragmenting regulatory map.
For researchers, educators, journalists and the public. A clear, trustworthy reference — and a shared language — for understanding and scrutinising empathic AI. Free to read, so it’s free to use, teach and cite.
Free to access
Unusually for a standard of this kind, IEEE 7014.1 is free to read. We believe a benchmark for technology this intimate should be open to the people it affects — so it can be cited in policy, taught in classrooms, and used by builders of every size, not just those who can afford it.
Who made it
IEEE 7014.1 is the work of the IEEE working group on Ethics & Emulated Empathy in Partner-Based GPAI — chaired by Professor Andrew McStay (Emotional AI Lab), with Ben Bland as vice chair and Karen Bennet as secretary — and was approved by the IEEE SA Standards Board in February 2026. The Emotional AI Lab, which has long studied empathy, emotion and AI, supported its development.
Call-to-action
Read it. Cite it. Build to it. IEEE 7014.1 is free, agreed by experts worldwide, and ready to use today.