ChatGPT is banned at your job. Your dyslexia adjustment isn't

Five major US banks now ban or limit ChatGPT at work. The list runs JPMorgan Chase, Goldman Sachs, Citi, Deutsche Bank, and Wells Fargo (company AI policy roundup, checked September 2026).

The Democratic National Committee barred ChatGPT and Claude for staff in April 2026, too.

It still allows Google Gemini for coding tasks. So the ban is never really about AI as a whole. It's about one tool, at one moment, for one stated reason.

If your dyslexia adjustment runs through a banned tool, this hits you differently than it hits everyone else. Most staff just switch apps and move on. You lose a fix for a specific, daily barrier.

You might be the analyst who runs a client email through ChatGPT first. It catches the sequencing errors dyslexia makes easy to miss. You might use Copilot to check a spreadsheet formula before it reaches a director.

Either way, the security memo was never written with your case in mind. It almost never says so directly. That silence is exactly why you need to raise it yourself, in writing.

Why companies are banning the exact tool you rely on

Most AI bans exist for one reason: data leaving the building. Almost 70% of employers restricting ChatGPT cite protecting confidential data (BlackBerry survey, via company AI policy trackers, checked September 2026). A bank worries about client data typed into a prompt.

A hospital worries about patient records ending up on an outside server. A law firm worries about privilege. None of that reasoning is really about you.

A dyslexic employee drafting an email with an AI tool is not the risk the policy was written for. But the policy rarely draws that line. It bans the tool for everyone, in one sweep.

Bans do get reversed once employers fix the underlying risk. Samsung banned ChatGPT after a 2023 data leak, then changed course in June 2026.

It gave roughly 125,000 staff enterprise ChatGPT access. It had built proper data controls first (company AI policy roundup, checked September 2026).

So what for you: the ban exists to stop data leaving the company, not to stop you working. That gap is the whole basis of your request.

Does your AI use actually count as an accommodation

Not every AI habit counts. The ADA and the Equality Act 2010 both use a simple test.

Does the tool remove a barrier tied to your dyslexia? Or is it just a general work habit anyone could use?

A worked example

A financial analyst uses ChatGPT two ways. First, to shorten long research notes into a quick brief. That's just a time-saver, and most colleagues could use it too.

Second, he runs a structure check on his own client emails before sending. This catches the transposed figures and out-of-order sentences that come from his dyslexia specifically. Only the second use is the real accommodation.

That distinction shapes your written request. Don't write "I use ChatGPT." Write "I use an AI check on outgoing client emails."

Then add why: "It catches dyslexia-related errors before a client sees them." That one line does most of the work.

So what for you: name the specific task the tool fixes, not the tool's brand name. A vague request is easy to refuse. A task-specific one is not.

The first lawsuit testing this exact conflict

Angeliz Cedeno wore Meta smart glasses at her security job at Walt Disney World. She had postpartum eye problems that made screens and glare difficult, and her doctor prescribed the glasses. Her manager later learned they were not ordinary eyewear.

Disney applied its ban on the devices and refused her accommodation request. Cedeno sued on 23 October 2025. Employment lawyers call it the first lawsuit of its kind.

It tests whether an AI-linked device can count as an ADA accommodation (case filed, reported October 2025). No court has ruled on the merits yet.

The case is still moving through the courts. It hasn't set a binding rule yet either way. But it confirms one thing: a company-wide tech ban is not an automatic shield against an individual accommodation claim.

So what for you: there's no final ruling to point to yet. But the legal framework treating your AI tool like any other accommodation request already exists.

What the law already says about blanket policies

You don't need a ChatGPT-specific ruling to see where this goes. UK case law already settled this for a different blanket rule. A tribunal found a security firm discriminated against a dyslexic guard.

It had refused him 15 to 20 minutes of leeway on shift starts. That rule applied to everyone (Mr RJ Bryce v Sentry Consulting Limited, judgment published February 2023).

A rule for the whole workforce is exactly what triggers the duty. It doesn't excuse it.

In the US, EEOC guidance from May 2022 makes the same point. An employer can be liable under the ADA if AI-based systems block a reasonable accommodation request. This applies whether or not the tool is company-approved.

	UK	US
Legal duty	Reasonable adjustments, Equality Act 2010	Reasonable accommodations, ADA
Who decides "reasonable"	Employer, tested by a tribunal	Employer, tested via the ADA process
Where to escalate	ACAS, then tribunal	EEOC, within 180 to 300 days
Damages if refused unfairly	No cap	Capped $50,000 to $300,000 by employer size

So what for you: "it's against policy" is not a legal answer, in either country. It's the start of a conversation your employer owes you.

What this means for you

Don't keep using a banned tool quietly and hope nobody checks the logs. That turns this into a disciplinary problem, not an accommodation one. It hands your employer a clean reason to act that has nothing to do with your dyslexia.

Ask for a narrow, written exception instead. Name the exact tool, the exact task, and how long you've used it. The reasonable adjustments builder turns that into a written request and a draft email in about two minutes.

If IT security is the real objection, ask about an enterprise-licensed version with the employer's own data controls. That's the same route Samsung eventually took. It can solve both problems at once.

If you're told no with nothing offered instead, that's the point to escalate. Under UK law, your employer must justify why no adjustment works, and explore alternatives first. Under US law, a flat refusal with no discussion is itself a warning sign for a later claim.

Four steps, in order. Write the request naming the exact task, not the tool's brand name. Send it as an email to HR or your manager, so there's a timestamp on it.

Ask directly whether an enterprise or on-device version solves the security concern. Keep a copy of the answer, even a verbal one you summarise back in writing.

So what for you: send the written request this week. Do it before the policy becomes the reason you're disciplined, not accommodated.

The ADA accommodations guide and the disclosure decision guide cover how to frame this if you haven't disclosed yet. UK readers funding an approved AI tool can check the Access to Work guide. The Tech Fund can cover software costs for anyone under six weeks into a role.