1 in 2 workers hide their AI use. Yours is protected by law.

Nearly half of US employees admit to hiding their AI use at work, and one in seven avoid telling their manager at all (Laserfiche survey of 1,000 US workers, via Reworked.co, August 2025). The top reason people gave for the discomfort: it feels like cutting corners. If you're dyslexic and you draft emails with Copilot, dictate reports with Dragon, or ask ChatGPT to tidy a paragraph before you hit send, you've probably felt that same discomfort.

This article explains why that discomfort is pointed at the wrong target. Assistive AI and dictation tools are not a favor from an understanding manager. Under UK and US disability law, they sit in the same legal category as a screen reader or extra time in an exam.

Why hiding feels safer than admitting it

This instinct to hide AI use starts as ordinary workplace anxiety, not something specific to dyslexia. Reworked's research found employees worry about looking lazy, unoriginal, or out of step with unwritten rules, and only one in three believe their workplace has a clear AI policy at all (Reworked.co, August 2025). Vague policy breeds guesswork, and guesswork breeds silence.

Perception bias compounds the problem. A 2026 study covered by Forbes found that when women and men submitted identical AI-assisted resumes, evaluators rated the women as less competent and credited the men with initiative (Forbes, June 2026). If a tool can be read as either a shortcut or a strength depending on who is holding it, the fear of judgment is rational, not paranoid.

For a dyslexic employee, that general workplace anxiety stacks on an older one: the worry that using a tool to write or read proves you can't really do the job unaided. That worry predates AI by decades. It's the same logic that once made people hide spellcheckers, and before that, calculators.

The doubt you feel about using ChatGPT at your desk is a documented, workplace-wide pattern, not a personal failing. That matters when you decide whether to keep your workaround quiet or put it on record.

The legal test does not care how you look

In the UK, Section 20 of the Equality Act 2010 requires an employer to make reasonable adjustments where a disability puts an employee at a substantial disadvantage. Section 21 makes a failure to do so unlawful. Assistive technology, including AI writing tools and dictation software, sits squarely inside the adjustments an employer is expected to consider.

In the US, the ADA's reasonable accommodation doctrine works the same way. An employer must provide or permit tools that let a disabled employee perform their job, unless doing so causes undue hardship. The Job Accommodation Network (JAN), the federally funded accommodation resource, lists speech-to-text and AI writing support as an established accommodation category, on the same list as screen readers and ergonomic keyboards (askjan.org).

Neither law asks whether you could theoretically manage without the tool. Neither requires you to prove you tried harder first, or tried a cheaper option, before reaching for the one that works.

If your employer questions why you "need" Copilot when a colleague doesn't, the honest answer is that the law does not measure your adjustment against anyone else's baseline. It only asks whether the tool addresses your disadvantage.

Who is already paying for this

In the UK, Access to Work already funds the tools behind this week's toolkit comparisons on this site. Dragon Professional Anywhere costs £600 plus VAT a year and can be covered in full for eligible applicants. Employees under six weeks in a new role can get AI writing tools and dictation software funded at 100% through the Access to Work Tech Fund, with no employer contribution required (DWP, 2026).

That funding exists because assistive AI is already classed as a legitimate adjustment, not because it's an experimental perk someone might claw back later. Read the site's Dragon vs Copilot comparison or the ChatGPT vs Copilot breakdown for the exact costs and what each tool handles best. In the US, JAN's accommodation directory lists comparable AI writing and dictation tools without treating them as a special case.

Standard Access to Work awards outside the Tech Fund window are still subject to the frozen £69,260 annual cap, and processing has been running to roughly 37 weeks this year. Neither of those numbers changes whether AI or dictation software counts as a genuine adjustment; they only affect how quickly the funding arrives. If your claim is delayed, your employer's own duty to make adjustments runs independently of Access to Work, so you can ask for the tool directly while the application is pending.

If you're paying for ChatGPT Plus or Dragon out of your own pocket while staying quiet about using it, you may be funding something in secret. Access to Work, or your employer's ADA duty, may already cover it.

What cheating would actually require

Academic and workplace guidance on generative AI usually targets one specific kind of dishonesty: presenting someone else's unaided work as your own, or breaking a stated rule about what's allowed on a given task. Neither applies to a dyslexic employee using dictation software to draft an email, or a text-to-speech tool to check a report before it goes out.

Disability guidance is explicit that assistive technology should never be conflated with academic or professional dishonesty. Cheating requires deception about a rule you agreed to follow. Disability law creates the opposite of a rule violation: it creates an entitlement.

A blanket "no AI" policy with no exception for a documented disability is itself the legal problem, not your use of the tool. Under both the Equality Act and the ADA, a rule that disadvantages a disabled employee has to be adjusted, not applied rigidly regardless of the cost to that employee.

A written record beats a guess

Picture two employees who both use Dragon to draft reports. One mentions it once in passing to a manager and otherwise stays quiet. The other emails HR: "I use Dragon Professional Anywhere to draft written reports accurately, as a reasonable adjustment for dyslexia."

Only the second employee has evidence if a later manager questions their methods, or if a report contains an error and someone asks how it was produced. The tool is identical in both cases. The paper trail is what changes the outcome.

If your employer has a company-wide AI ban, the fix is not to keep working around it quietly. It's to request a written exception as a reasonable adjustment, on record.

What this means for you

If you're already using Copilot, ChatGPT, or Dragon and keeping it to yourself, the safer move is to put it in writing as a reasonable adjustment rather than hope nobody asks. A written record protects you later if your output is ever questioned, the same way documented adjustments strengthened the claimant's case in the £470,000 Lloyds tribunal ruling.

If you haven't asked yet, start with a short request that names the task, not a diagnosis grade: "I use dictation software to draft client emails accurately." The reasonable adjustments builder turns that into a conversation plan and a draft email in about two minutes.

If your employer's policy bans AI outright, check your Access to Work eligibility before you pay for a workaround yourself. Funding is often available, and hiding the need for it isn't a step you have to take.

The fear of looking incompetent has kept nearly half the workforce quiet about tools they're already using. For dyslexic employees, the law settled the underlying question years ago. What's left is paperwork, not permission.