A Report on How Disabled and Deaf Artists in BC Are Using AI Tools
grunt gallery, 2026
This research and report was facilitated, written, and narrated by Kay Slater
Funded by the Canada Council for the Arts through the SEED grant
Accessibility Statement: This report is available in HTML, plain text markdown, audio with visual descriptions of any charts, and ASL video with a Deaf interpreter
Report licence: Creative Commons Attribution-ShareAlike 4.0 International
This licence requires that reusers give credit to the creators. It allows reusers to distribute, remix, adapt, and build upon the material in any medium or format, even for commercial purposes. If others remix, adapt, or build upon the material (adaptations), they must licence the modified material under the same terms. Credit must be given to Kay Slater and grunt gallery, the creator.
This licence only applies to the text and audio versions of the report. The ASL interpretation is not available for remix or redistribution and must stay within the original format and site.
Acknowledgements (Kay Slater): This report was funded by the Canada Council for the Arts through the SEED grant. A survey was completed by disabled and Deaf artists in southern BC. I am grateful for grunt gallery’s support and to every community member who gave their time and trusted me with their perspectives. I acknowledge my biases as a white, ambulatory, hard of hearing Canadian over 40 who is employed and not receiving disability support. I have tried to be clear when sharing my perspectives versus the data I collected, but I encourage readers to question me, talk to others, and keep collecting information.
Table of Contents:
- Section 1. Introduction: Who Wrote This and Why
- Section 2. Defining AI: What Do We Even Mean?
- Section 3. How AI Is Actually Being Used.
- Section 4. AI as Access Tool, Not Shortcut
- Section 5. Human Support: Essential and Structurally Limited
- Section 6. Risks, Fears, and Real Dangers.
- Section 7. Economic Survival and Ethics.
- Section 8. Shame, Secrecy, and Community Dynamics.
- Section 9. Copyright, Consent, and Contradiction.
- Section 10. What Would Need to Change
- Section 11. Key Findings Summary
- Section 12. Calls to Action.
- ASL Video: Final Note
- Appendix: Methodological Notes
- Appedix: Template letter (open new tab)
- Appendix: Survey research questions (open new tab)
Section 1. Introduction: Who Wrote This and Why.
I’m Kay Slater. I am a working artist and arts worker, not a scholar. I’m the exhibitions and accessibility manager at grunt gallery in Vancouver and a multidisciplinary artist whose practice centres on access and non-verbal participation. I use AI most often for auto-captions in meetings, which is essential to my career and helps me manage hearing fatigue.
This report documents what I learned over a year of research into how disabled and Deaf artists in BC are actually using AI tools and not just generative AI and LLMs like ChatGPT, but the full range of AI technologies that are already, or are in the process of being, integrated into tools our sector has used for decades. I wanted to understand what was working, what concerned people (without anyone feeling pressured toward the “right” answer), and where both the possibilities and the fears lie.
After six months of research, workshops, casual conversations, and attending lectures and conferences, I authored a survey of 44 questions. These were loaded into accessiblesurveys.com and offered to 100 artists across British Columbia who publicly identified as disabled or Deaf. I received 26 completed surveys. Forty-eight percent (48%) of those who responded were aged 30–39. Twenty-four percent (24%) were 50+. Twenty percent (20%) were 40–49, and eight percent (8%) were under the age of 30. I specifically sought artists or writers with an established creative practice of at least two years. Unless marked as my opinion or externally cited, the data in this report comes from that survey and paid follow-up conversations. The full survey is available in the appendices and is shared under the Creative Commons licence stated above.
What this is not:
This is not an academic study. It is not comprehensive, and it is not neutral. The sample is small and geographically limited to BC. I deliberately avoided contacting artists from Toronto and Montréal, as most of the studies available to me were already based there. I regret that this report excludes most of Turtle Island, and I invite any artist-researchers interested in continuing this work to reach out. This report is a starting point, not a conclusion.
Why it matters:
Disabled and Deaf artists’ voices are frequently missing from conversations about AI and the arts. They are particularly vulnerable to the real risks being discussed by Canadian artists: privacy concerns, lack of regulation, environmental impact, and the threat that organizations and governments will use AI as justification to defund human support roles. For those who have gained access through AI tools, there is an added dimension to that fear: if these tools (imperfect as they are) get counted as “enough,” will that erase the opportunity to ask for and receive human services?
This research does not pretend AI isn’t risky and it tries not to position AI as inevitable. It is about understanding how AI is actually being used so that disabled artists can be part of conversations about regulation, limits, and possibilities.
On scope and prior work:
This research was partly motivated by the national consultation on copyright and generative AI conducted by the Government of Canada in 2023, and the important work that CARFAC Ontario contributed to that process on behalf of visual artists. I wanted to focus specifically on Disabled and Deaf artists, and artists with disabilities, on the West Coast, whose perspectives were underrepresented in that conversation. For those interested in the broader national picture and one more focused on generative AI, the CARFAC recommendations and the Government’s “What We Heard (PDF)” report are valuable companion documents. (See: carfac.ca and ised-isde.canada.ca)
Section 2. Defining AI: What Do We Even Mean?
The working definition of “AI” used in this research: Computer systems (hardware and software) that perform tasks typically associated with human intelligence, such as learning from data, recognizing patterns, making predictions, and generating outputs. This includes machine learning, automation, and generative systems.
Survey responses revealed that understanding of AI varied widely:
- ~30% reported a strong technical understanding of AI
- ~30% provided broadly accurate but simplified explanations
- ~20% associated AI primarily with generative tools like ChatGPT
- ~15% demonstrated misunderstandings (describing AI as autonomous or conscious)
- ~15% expressed uncertainty or reflected on the ambiguity of the term
“Artificial intelligence, or AI, is an umbrella term for computer programs and systems that can perform tasks historically associated with human intelligence (e.g. reasoning, learning, acting autonomously).” — Anonymous, disabled BC artist
“Artificial Intelligence is a term that I think is intentionally obscure… But AI can refer to programs like ChatGPT and Midjourney, huge generative AI projects that are more commonly what comes to mind when people say AI.” — Autistic artist living with OCD, PTSD, and hypermobility.
Most respondents show a working understanding of AI in relation to their own use. Definitions were shaped by personal experience and media narratives, but not by ignorance.
[Kay’s observation:]
The most common “mistake” respondents made was conflating AI with generative AI. This is not a knowledge failure; it reflects how the media and the arts sector have framed these conversations. The framing actively harms disabled artists who depend on non-generative AI tools because it lumps essential access tools in with contested creative ones.
Section 3. How AI Is Actually Being Used.
Despite varied definitions, AI use was nearly universal. Over ninety percent (90%) of respondents self-reported using AI-powered tools regularly or occasionally, with more than half using them regularly.
Fig 1 Chart description: A two-column chart with header use pattern on the left and percentage reporter on the right.
| Use Pattern | % |
|---|---|
| Yes, regularly | 57.7% |
| Yes, occasionally | 34.6% |
| Tried but stopped | 7.7% |
| Not sure | 3.8% |
Note: some respondents chose not sure and another option.
The most widely used tools were not generative AI. They were assistive technologies.
- Accessibility and language tools (voice-to-text, auto-captioning, transcription, text-to-speech, screen readers, translation): ~70–85% of respondents
- Platform and embedded AI (navigation, predictive text, social media algorithms, voice assistants): ~60–75%
- Core LLM tools (ChatGPT, Gemini, Claude, Copilot): ~55–65%
- Vision and sensory AI (Be My Eyes, OCR, AI hearing aids, sound detection): ~30–45%
- Creative and media tools (image generators, audio/video editing, voice cloning): ~20–35%
While many people reported using LLMs, they were quick to identify their use as primarily for writing support, admin, and brainstorming.
“I recently bought a new pair of hearing aids that incorporate AI based off of thousands of speech samples. It is fascinating to learn how that works. I am interested in creating work based off of this concept- is AI hearing for me?”— Anonymous, disabled artist, 30+
[Kay’s observation:]
The low use of creative tools is notable. It reflects both the ethical concerns about generative AI and the fact that, for most respondents, AI’s primary value is in access and administration, not creative production. This is a theme repeated throughout this report.
I was surprised that so few artists listed photo and audio editing tools, since many are AI-powered. This might confirm my suspicion that most people don’t recognize embedded technology in their familiar tools as AI at all. The shame and debate are concentrated on generative tools, while “invisible” AI (automation) continues to be used daily and uncontested. In conversation around behavioural targeting of advertisements, social media and product suggestion algorithms, predictive text when messaging or doing search, and voice assistants, I often found people startled to be reminded of the AI integration. I had interesting conversations with folks who first objected to these things being driven by AI, and in some cases, they were right that the technology had evolved from rule-based, statistical models. However, over time, these familiar tools now use machine learning and AI-enhanced summaries; most notability in these conversations was the evolution of search engines. In one conversation, someone was visibly shaken that they hadn’t noticed that the contents they had been reading were AI summaries.
Section 4. AI as Access Tool, Not Shortcut
The strongest and most consistent finding of this research: for the surveyed disabled and Deaf artists, AI is primarily an access tool used to fill gaps left by unavailable, unaffordable, or inconsistent human services.
Almost all of respondents (90%) reported using AI tools to access services that are unavailable or difficult to obtain through human support.
Why respondents use AI instead of human services:
Fig 2 Chart. Description: A two-column chart titled “Why…” with header “reason” on the left and “percentage of respondents} on the right.
| Reason | % of respondents |
|---|---|
| Human services not available when needed | 85% |
| Human services too expensive | 77% |
| AI is faster | 65% |
| AI provides independence / privacy | 62% |
| Human service wait times too long | 62% |
| AI is “good enough” | 58% |
| Prefer hybrid use | 42% |
| Human services not available in location | 27% |
When rating quality compared to human services: Thirty-six percent (36%) said quality varies by tool and thirty-two percent (32%) said AI is worse but still useful. Eight percent (8%) said AI is better; and 8% reported they cannot compare because they have no human services available at all. For some respondents, it is not an alternative. It is the only available form of access.
“I cannot have a human to stand by all 24 hours.” — Anonymous, disabled BC artist, 40+
“Disabled artists are likely going to use AI for access support because it is helpful, but we would much rather be relying on humans instead. I would much rather go to a human being with my questions, it’s just a matter of having access to those humans and them being paid well.” — Anonymous, disabled BC artist, housebound with complex chronic illness, 30+
“We wouldn’t need AI if you listened to us the first time.” — Anonymous, Indigenous Trans woman and artist with neurodivergence, 30+
[Kay’s observation:]
When humans use digital tools to supplement our work, I have been calling this “augmented intelligence”, a term with a longer history than the current AI conversation. (See: Douglas Engelbart’s Augmenting Human Intellect: A Conceptual Framework, 1962; W. Ross Ashby’s Intelligence Amplification (PDF); J.C.R. Licklider’s Man-Computer Symbiosis.) I came across the term when I read a report (more than half a decade ago) by IBM that argued that AI should be seen as a supportive tool in sectors like business and healthcare, not as a replacement.
The key distinction that keeps getting lost in the broader arts conversation is this: using auto-captions because a captioner was not budgeted for is not the same as using an image generator to produce work. Also, it’s worth considering that the organization that didn’t budget for a captioner probably wouldn’t have provided captions before auto-captions existed. I don’t think the fight should be focused solely on the tool. It should be about getting “access” on the budget in the first place.
These are being collapsed into one conversation, and that collapse harms disabled artists.
Section 5. Human Support: Essential and Structurally Limited
Respondents were clear that human support workers are preferred, and that systemic barriers, not personal preferences, drive AI use.
What respondents said human support workers provide that AI cannot:
- Greater accuracy and contextual understanding
- Nuance, cultural knowledge, and lived experience
- Relational care, warmth, and trust
- The ability to respond to individual and situational needs
Why human services remain out of reach:
- Expensive and requires advance scheduling
- Not available outside 9–5 hours
- Require coordination that adds a labour burden to the person seeking access
- Scarcity. There is a deficit of trained support workers.
“Human access is built on monday to friday, 9–5, which is not accessible.” — Anonymous, Indigenous Trans woman and artist with neurodivegence, 30+
What AI fails at:
- Inaccuracy, hallucination and confident errors that require human verification (~80–90%)
- Lack of cultural, contextual, and artistic nuance (~70–80%)
- Linguistic and racial bias, particularly for non-English speakers, accented speech, AAVE, and Indigenous languages (~60–70%)
- Accessibility failures within accessibility tools—captions that are inaccurate, setup too complex, not screen reader friendly (~60–70%)
- Cognitive overload and steep learning curves (~50–60%)
“Disabled artists are likely going to use AI for access support because it is helpful, but we would prefer to be relying on humans instead.” — Anonymous, disabled BC artist, housebound with complex chronic illness, 30+
[Kay’s observation:]
Almost every failure around captioning accuracy that respondents reported came from non-dominant language use: accents, non-English speech and cultural speech patterns. This is consistently framed as AI’s failure, but there are human decisions behind these systems. If the people designing and training these tools are not disabled, not multilingual, and not from racialized communities, the product will reflect that. The disability community is not the target market for most of these tools, which means our needs remain an afterthought.
The irony is that, during the initial three years of pandemic distancing, the planet contributed enormous amounts of speech data to the models on which these tools are now built. The tools got better through our collective use, but how many of us knew our data was being recorded? How many administrators understood that the digital tools they pivoted to were feeding data centres, often not subject to local privacy laws? This is precisely why regulation matters: not to slow down access, but to ensure the people whose data builds these systems have some say in how it’s used. I hope a major take away from this report is that we need to make an effort to push national and provincial regulations because, independently, users and organizations are not always equipped to understand the outcome of their tool use decisions.
(See: Feng et al., “Towards Inclusive Automatic Speech Recognition,” Computer Speech & Language, Vol. 84, March 2024; Guo et al., “Bias in Large Language Models: Origin, Evaluation, and Mitigation,” George Washington University, November 2024.)
Section 6. Risks, Fears, and Real Dangers.
Respondents expressed consistent concern across multiple dimensions of AI. Most selected 8–12 concerns each. Apprehension is layered rather than focused on a single issue.
Fig 3 Chart. Description: A two-column chart titled top concerns, with header concern on the left and percentage on the right.
Top concerns:
| Concern | % |
|---|---|
| Privacy and data collection | 92% |
| Lack of regulation in Canada | 88% |
| Environmental impact | 84% |
| Bias in AI systems | 84% |
| Accuracy and reliability | 80% |
| Job loss for human workers | 80% |
| AI used to cut human services | 80% |
| Losing skills / dependency | 76% |
| AI flooding creative industries | 76% |
| Copyright / IP concerns | 72% |
The dominant fear specific to disability access (~70–80%): AI will be used to justify the reduction or removal of human-provided accessibility services. Access will be declared as “solved” and then funding will be removed.
Other major concerns included:
- Standardization erasing individual needs and relational forms of care (~60–70%)
- Bias, exclusion, and lack of disabled leadership in AI development (~60%)
- Access becoming privatized and paywalled (~40–50%)
- Loss of human connection and community care (~30–40%)
- Surveillance, data misuse, healthcare discrimination, and insurance denial (~30–40%)
- AI tools that are themselves inaccessible (~20–30%)
“It does not empower us in real time. Individual and societal transparency, open dialogue and requests for consent make us feel safe. This is not what is happening right now… ask us if it’s ok to use AI before proceeding with it.” — Anonymous, Culturally Deaf, Sign Language Speaker (ASL & LSQ), Female, Queer Artist, 50+
Participants do not reject AI because they don’t understand it. They are using it despite understanding its risks.
[Kay’s observation:]
What struck me reading these responses is that these artists are simultaneously using these tools and holding well-articulated, consistent fears about them. I think this comes from survival under capitalism plus ableism. You use what is available while knowing it is not safe, not enough, and not what you asked for. This is the same logic that leads someone to use a shuddering elevator because there are no ramps, and then only report the elevator as a hazard on the way out.
Section 7. Economic Survival and Ethics.
Many disabled artists are navigating significant financial and structural constraints alongside these decisions.
- 72% of respondents identified grants as a primary income source, not supplemental
- 68% relied on commissions and projects; 60% on teaching
- 43% reported receiving disability benefits, with 26% opting not to disclose their status.
- Disability benefits both provide necessary stability and impose income thresholds that complicate or restrict additional earnings from art
Within this context, AI use cannot be understood as a purely ethical or aesthetic choice. It is shaped by poverty, time constraints, benefit restrictions, and limited access to support.
Two distinct groups emerged from the data:
Group A – Pragmatic survival users: Use AI when needed, accept contradictions, and prioritize access over ideological consistency.
Group B – Ethical resistors: Avoid AI (especially generative tools), maintain strict ethical boundaries, accept the financial, access, and creative limitations that follow.
“This is an area I struggle in all the time, and the only way I know how to navigate it is by staying true to myself and trusting that I am operating in a way that reflects my values and ethics. I don’t want to compromise those things even if it means I make more money – that is not how I want to live my life.” — Anonymous, disabled, agender queer bisexual Indigenous person with AuDHD, Ehlers Danlos Syndrome, and anxiety disorders, 30+ ; on being expected to weigh ethics against access survival
[Kay’s observation:]
Neither group is wrong. Both exist under conditions of “structural failure” (capitalism + ableism). Two moments from this year have stayed with me and talked about why this tension is so difficult to resolve.
The first was a response by a racialized, disabled artist who stated that the reason they were so against AI in principle was that their culture and knowledge had been stolen for hundreds of years (and were still being appropriated to this day). There was no wiggle room for being OK with anyone using their work, regardless of another person’s access needs. Someone else’s barriers did not and could not negate their right to fiercely protect their intellectual property, especially after the constant labour of proving their work’s value in predominantly white arts spaces. Their position is not a failure of empathy. It is the product of a different but equally valid set of survival conditions.
The second was a conversation I had with a racialized, partially deaf, disabled artist who is blind. They described how the internet and AI had given them access to things they had never had before: the ability to encounter images shared online, to “visit” cultural spaces they couldn’t physically or financially reach, to get information without depending on another person’s availability or willingness. They had been using and adapting content from Wikipedia for creative projects for years. For them, the adoption of AI tools simply extended access that was already imperfect. They were frustrated by sighted artists who criticized AI for “stealing” content because until those artists were in a position where looking at any work required consent from another person, they couldn’t fully understand what access had always cost people like them. Even their AI-powered glasses were a source of ethical unease because the glasses filmed people in order to provide navigation and description, trading one person’s privacy for another’s access. They didn’t let it keep them up at night.
My point is not that one position is wrong or better. It is that these are genuinely incompatible lived realities navigating the same broken systems. The questions I asked and the data I received did not resolve this tension or answer the question of whether using AI tools is ethical. Every respondent was firmly in one group or the other, and named it as a non-negotiable daily reality.
Section 8. Shame, Secrecy, and Community Dynamics.
[Kay’s observation:]
One of the most significant and ongoing trends from this year was how much shame surrounds AI use among disabled artists.
In conversation and in survey responses, AI use was frequently hidden. People would say AI was wrong, then describe using it. Often if the conversation was shut down, it wasn’t due to ignorance. It was protective. In smaller disability and Deaf arts communities, the stakes of a “wrong” or unpopular opinion are high. Losing access to community risks losing a network of care. The talk is sharper, more laced with trauma, distrust, and finger-pointing.
“For some disabled artists, AI is less about novelty and more about access. I use AI as a processing and communication support, to slow my thinking down, organize language, and reduce overwhelm, especially in contexts where I am expected to respond quickly or articulate complex ideas on demand. What has been challenging is that negative perceptions of AI often lead to judgment or policing of its use, rather than curiosity about how it can function as an accessibility tool. That dynamic can pressure disabled people to hide their use of AI or prove their legitimacy, instead of being supported in using tools that increase clarity, participation, and agency.” — Anonymous, disabled BC Artist with disabilities that include: mobility challenges, severe allergies, sleep disorder, anxiety, depression, sensory hypersensitivity leading to cognitive overload, 50+
The conditions for shame are real. Being seen using a tool the community has labelled as evil can mean being cut out. And for many disabled artists, cutting out means isolation. Not just social, but from the material networks that provide care and survival.
Part of what I wanted this project to do was to be seen, publicly, as someone who will not shame others for using these tools. To create space in community for conversations that don’t require everyone to hold the same values in order to receive care or support. The polarization around AI replicates a wider pattern I’ve observed in disability community organizing: that disagreement can become grounds for the withdrawal of mutual aid.
Section 9. Copyright, Consent, and Contradiction.
On AI training and copyright:
- 80% stated AI should not train on artists’ work without permission or compensation
- 80% agreed copyright law needs to be fundamentally rethought for the AI age
- 24% described their perspective as “complicated”
- 20% felt that if AI is trained on public content, outputs should be freely usable
- 8% said they don’t know enough to have an opinion
On their own practice:
- ~50–60% described their own practice as ethically grounded (never using others’ work without permission)
- ~25–30% acknowledged using others’ work with justification (fair dealing, transformative, or not knowing better)
- ~20–30% expressed uncertainty about copyright and licensing norms.
On AI vs. human creative work in funding contexts:
- ~60–70% felt AI and non-AI work should not be judged together
- ~60–70% called for mandatory disclosure and transparency
- ~50–60% identified fairness concerns related to labour and effort
“Arts organizations should not judge disabled artists too quickly for using AI; AI can be a game-changer for access and should not be treated as lesser or illegitimate.” — Anonymous, disabled, legally blind, dyslexic BC artist living with ADHD, chronic pain, chronic illness, 40+ ; on being expected to weigh IP ethics against access survival
[Kay’s observation:]
The artists I spoke with and who responded to this survey hold their own practices to a different ethical standard than they hold AI systems, and I hesitate to call this hypocrisy. It reflects a gap between what people do online out of habit and what they say they believe when they stop and think about it. People operate inside messy, normalized systems but articulate strong ethics when asked directly.
The same dynamic appeared to me at a grant selection panel: a panel member sneered at em dashes as AI-generated and wanted to dismiss the application. The moderator had to step in to say that AI use in disability arts grant applications for their organization is an active access conversation and that it hadn’t been forbidden in the rules for the application. We spent a considerable amount of time talking about prejudice around AI and had to take a break before we returned to the application. Ultimately, the other juror recused themselves, citing that they were too angry to look at the application objectively.
Section 10. What Would Need to Change
When asked what would need to be true for them to feel better about AI in relation to disability access, responses were structural, not technological:
Fig 4 Chart. Description: A two-column chart with header condition on the left and percentage on the right.
| Condition | % |
|---|---|
| Better privacy protections and regulation | 96% |
| AI companies being transparent about how tools work | 92% |
| Guarantee AI won’t replace funding for human services | 92% |
| More disabled people involved in AI development | 88% |
| More education about disability for AI developers | 84% |
| More education about AI for disabled artists | 80% |
| Better accuracy and quality of AI tools | 76% |
| More accessible AI tools themselves | 76% |
| Lower costs or free access to AI tools | 72% |
This is a [conditions for trust] dataset, not a feature request. Respondents want regulation, transparency, and protection of human services before they want better tools. The data is clear: trust in AI will depend on how it is governed and integrated, not on technological advancement alone.
When asked how AI should be used by arts organizations:
- ~70–80% said AI must be optional, consent-based, and user-controlled. Access should not be contingent on engaging with AI
- ~70–80% said AI should supplement human support, not replace it
- ~60–70% called for transparency and disclosure
- ~60–70% rejected generative AI in artistic and administrative contexts
“AI can be helpful as an access and support tool, assisting with drafting, organizing ideas, and reducing physical or cognitive strain in the creative process. My boundary is that it should not replace artistic intent, voice, or decision-making. I also want to name that boundaries are often shaped by stigma, negative perceptions of AI can create pressure to hide its use or to over-police work so nothing appears ‘too AI,’ which undermines its legitimacy as an accessibility tool.” — Anonymous, disabled BC Artist with disabilities that include: mobility challenges, severe allergies, sleep disorder, anxiety, depression, sensory hypersensitivity leading to cognitive overload, 50+
Section 11. Key Findings Summary
- AI use is widespread.
Over 90% of respondents are already using AI tools regularly or occasionally, primarily accessibility technologies rather than generative ones. - Understanding is low and shaped by media.
Most people understand AI through the lens of generative tools. Non-generative AI (embedded in tools used daily) is largely unrecognized, or used without naming it as AI. - AI can be an access tool.
For most respondents, AI fills gaps in services that are unaffordable, unavailable, or untimely. This is a structural problem, not a preference. - Human support is preferred, but structurally inaccessible.
The issue is not humans versus AI. The issue is access to humans. - Generative AI is widely distrusted.
There was near-universal concern about training data, consent, and the ethics of generative AI. Most respondents draw a firm line here. - Shame is a serious barrier.
Many disabled artists hide their AI use due to fear of community judgment, which prevents honest conversation and support. - Fears are structural, not imaginary.
The highest concern is that AI will be used to justify cutting human services. This is already happening. - Conditions for trust are clear.
Respondents want regulation, transparency, and guarantees that AI will not displace human support before they want better features.
“We aren’t your gotcha argument or whataboutism when it comes to AI-generated art. I, for one, just want my captions, and didn’t sign up for uncanny, mud-tinted Ghibli clones everywhere.” — Anonymous, autistic BC Artist living with OCD, PTSD and hypermobility, 25+
Section 12. Calls to Action.
For individuals
- Learn what tools you are using and whether they are AI-powered, including the ones you already trust.
- Read the conditions of use if you are worried about data, privacy, and information use.
- Reduce the shame you direct at others for using tools you are not comfortable with.
- If you have capacity, advocate for regulation rather than policing each other. Write an email to your boss or board, write a letter to your local government, and ask what they are doing about AI policy.
For arts organizations
- Do not implement AI tools without the consent of those they affect. Consent is easier than you think – it’s slowing down that is hard.
- Do not use AI to justify reducing or eliminating human access roles: captioners, interpreters, support workers – especially if you weren’t funding them before.
- Develop clear AI policies that distinguish between generative and access tools.
- Include disabled artists in policy development, not after the fact.
- Fund professional development on digital tools, including AI.
- Do not judge disabled artists for AI use without understanding why and how they are using it. Allow for nuance and individual access needs. Equity is complicated.
- Don’t assume someone is using AI. Assumptions significantly impact access to funding and support, so they must be taken seriously. Talk to your artists, applicants, and grantees.
For funders and government
- Regulate AI use in Canada with urgency, especially data privacy and the protection of user data.
- Fund human access services so that AI is a supplement, not a survival strategy.
- Support disabled artists’ participation in AI governance and policy conversations.
- Require disclosure and consent in any AI-assisted programming or application processes. Be clear on what you accept, what you don’t accept, and how you are making decisions.
ASL Video – final note:
The appendix is available online with methodological notes, a template call to action letter you can use to send to your local government or employer governance teams, as well as the full list of survey questions used to generate this survey. All of these assets as well as the text and audio report are shared under the CC BY 4.0 licence and can be reused with attribution. The ASL video is not licenced under this share alike licence and cannot be duplicated or remixed.
Accessibility provider: ASL Interpreting Inc.
Translation provided by: Pamela Witcher
[ASL Video Interpretation Ends]
Appendix: Methodological Notes
Conversations that were held prior to the survey and calls for participation provided insight and focus for the final report, helped shape the research questions themselves and offered a list of technologies to be tried in gallery before the survey went out. Conversations were held with Deaf and disabled artists and creative artists with disabilities from across BC.
Surveys were collected using accessiblesurvey.com. When respondents could not use the online system, they were offered the survey in audio, video, plain text, or as a Word document. Respondents were compensated $150 regardless of survey completion, with $100 in access fees provided for any assistants. Respondents were offered the opportunity to review the survey and choose whether to be identified or anonymized.
An unanticipated limitation: the survey platform did not collect names, making it impossible to connect identity disclosure preferences to specific responses. All respondents were therefore anonymized unless explicit consent was confirmed before publishing.
A small amount of data (less than 3%) was lost when respondents expanded on multiple choice responses and the system recorded only that they respondent had opted to share more. This is acknowledged as a gap in the data.
Sample size is small and geographically limited to BC. This research should be read as a starting point and community document, not a statistically representative study.