1. The Story Behind the Day
A new international observance
Unlike many calendar days with decades of history, this observance is genuinely new. The General Assembly adopted A/RES/80/302 on 31 August 2026 and established the same date as an annual international day beginning in 2027.
The date connects contemporary AI governance to a historical milestone: 31 August 1955, when the proposal for the Dartmouth Summer Research Project on artificial intelligence was put forward. That proposal helped establish artificial intelligence as a named research field.
The 80th-session resolution record and the General Assembly plenary transcript support the first-edition timing. Calendar 5.0 therefore treats 2027 as the inaugural observance rather than back-projecting an earlier annual tradition.
Why “safe, secure and trustworthy” matters
AI systems increasingly influence communication, education, employment, healthcare, public services and information access. That creates potential benefits, but also risks involving bias, opacity, privacy, misinformation and unequal access.
The UN framing is deliberately broad. It concerns international cooperation and the conditions under which AI can be developed and used responsibly.
The first edition has no established tradition yet
Because 2027 is the first annual observance, there is no long history of rituals, campaign symbols or recurring public events to describe. The honest editorial approach is to explain the formal basis and the questions the day creates without inventing a tradition that does not yet exist.
2. What Can We Learn?
Technology can widen or narrow participation
AI tools can help with summarisation, translation, drafting, speech-to-text, planning or alternative communication. Those uses may reduce barriers for some disabled and neurodivergent people.
The same systems can exclude people if training data, interfaces or evaluation criteria reflect narrow assumptions.
“Trustworthy” needs to mean something operational
A trustworthy system should not merely carry a reassuring label. People need to know when AI is being used, what it can influence, how errors are corrected and how human review works.
Trust depends on accountability, not only technical performance.
Access and power are linked
The benefits of AI are not distributed automatically. Some communities have better connectivity, language support, computing resources and influence over design decisions.
An AI divide can therefore mirror existing social and digital divides.
3. What Needs to Change?
Include affected users before deployment
People who will be evaluated, assisted or affected by an AI system should have meaningful opportunities to shape requirements and identify risks.
This includes disabled and neurodivergent users whose behaviour may not match the patterns assumed by designers.
Build clear challenge routes
If an AI-supported system affects access to education, work, benefits or healthcare, users should be able to understand outcomes and challenge mistakes.
A human review process should be real, not a decorative button that sends people back into the same automated loop.
Test across languages and communication styles
AI systems often perform unevenly across languages, accents and forms of expression. Evaluation should include diverse users and realistic contexts rather than only benchmark data.
Accessibility testing should also cover cognitive load, clarity and sensory design.
Protect privacy and agency
Helpful AI can become intrusive when people cannot choose how their data is used. Organisations need clear purpose limits, meaningful consent where appropriate and proportionate data collection.
Procurement can lock in risk before users see the system
Many organisations do not build AI themselves; they buy tools from vendors. That makes procurement a major governance point. Contracts can require accessibility evidence, security standards, audit rights, data-use limits and clear responsibilities when a system produces harmful errors.
Buying a product should not mean outsourcing accountability. The organisation using the system still decides where it is appropriate and how people can challenge its outputs.
AI literacy should include uncertainty
People benefit from understanding not only what AI can do but where it can fail. Generated text can sound confident while being wrong, and automated classifications can hide uncertainty behind a score.
Teaching users to verify important outputs, recognise limits and preserve human judgement is part of trustworthy use.
Accessibility features should not become compulsory surveillance
AI can support captions, transcription, reading assistance or personalised interfaces. Those benefits should not require users to surrender unnecessary personal data or accept constant monitoring.
Accessibility and privacy are not opposing goals. Good design can minimise data collection, explain processing clearly and let users choose when an AI feature is active.
Organisations should also plan for failure. When an AI tool is unavailable, wrong or inaccessible, people need a workable alternative. A trustworthy system includes graceful fallback rather than making one automated route the only way to receive a service.
4. What Can You Do?
Take Part — Use AI deliberately rather than automatically
Try tools where they genuinely reduce friction, but notice what information you share and verify important outputs. If a tool helps you communicate, organise or access information, that can be valuable; you do not need to adopt every new system simply because it is marketed as intelligent.
Show Your Support — Ask who is missing from AI design
When discussing new systems, raise questions about disability access, language, bias, privacy and appeal routes. Amplify researchers and community advocates who examine real-world effects instead of treating AI inclusion as a purely technical problem.
For Organisations — Require an impact and accessibility check before high-stakes use
Document what an AI system does, who can be affected, what data it uses, how errors are challenged and where human judgement remains. Include disabled and neurodivergent users in testing, and do not deploy a system simply because a vendor calls it trustworthy.
5. Explore the Movement
The UN observance page provides the emerging international framework. The General Assembly resolution list and plenary record establish that annual observance begins in 2027.
Related Calendar 5.0 entries include World Telecommunication and Information Society Day, International Day for Universal Access to Information and observances about education, digital learning and public institutions.
No 2027 programme or theme beyond the formal purpose is asserted before the UN publishes one.
Go to the people behind the work
Start with United Nations · Safe and Trustworthy AI Day. Follow their own information and work; this source does not confirm a 2027 programme.
Explore this source ↗More information & organisations
Follow the original initiative wherever it can be identified. The research reference remains available for transparency.
Official observance information available · 2027 programme not verified
6. Why This Day Matters to the Neurodiversity Foundation
This new AI day matters directly to NDF because neurodivergent people may encounter AI both as an accessibility tool and as a system that judges them.
A writing assistant can reduce executive-function load; an automated recruitment screen can misread unconventional communication. Both are examples of AI, but they create very different distributions of power.
NDF supports technology that expands agency rather than forcing people toward one behavioural norm. That requires transparency, user choice, human appeal and real participation in design.
For us, the first 2027 observance is an opportunity to insist that “safe, secure and trustworthy” includes people whose minds, communication styles and support needs differ from the assumed default.