SD AI StudioLights On
San Diego's shadow AI amnesty

Turn the lights on.

Most people using AI at work are doing it on personal accounts, keeping quiet about how good they've gotten. The research calls them secret cyborgs. Really, everyone's just figuring out what to do and how. Show the workflow. Keep the credit.

The Impact Lab asked: who decides the rules?

The people already doing the work. The rules should start from what they built.

Part one · The 30-second tier test

Is the thing you're doing with AI risky?

Five questions. No login, and this part never leaves the page. It's the same test enterprises use to tier AI work, translated into plain language. Green means go. Yellow means a second set of eyes. Red means slow down, on purpose.

01What data goes into the tool?
02Who is affected if the output is wrong?
03How visible is the output?
04Is the action reversible?
05What does being wrong cost?
Part two · Why the hiding happens

Shadow AI is signal, not sin.

Nobody's doing anything wrong. Most people just don't know what's going on yet, and everyone is figuring out what to do and how. The numbers say quiet AI use is everywhere and it works.

The gains

MIT found 95% of official corporate AI pilots produce no measurable return while people's personal workflows quietly deliver. The best AI deployment in most buildings is the one nobody sanctioned. It works because the person who built it redesigned their own job around it.

The fear

Reveal that a three-hour task now takes thirty minutes and you don't get an afternoon back. You get six more tasks, and fast becomes the silent new baseline. Nearly half of workers say admitting AI use to their manager feels like looking lazy or less competent. Keeping quiet is a rational read of a bad incentive.

Bans make it worse

46% of workers say they'd keep using AI through an explicit ban. Blocking pushes the work onto personal phones and home machines where visibility drops to zero. A ban without a better path just moves the risk out of sight.

Amnesty flips it

When disclosure is framed as capability mapping instead of audit, and nobody gets punished for showing their work, disclosure rises 60 to 70%. Credit one builder publicly and every quiet builder recalculates. Punish one and everyone keeps quiet for good.

"Your deployment already happened. Your only decision is whether you get to see it."

David Shapiro, Shadow AI: The Definitive Report (July 2026), the one-line case for amnesty
Part three · The San Diego shadow AI census

Stand up and be counted. Anonymously.

Twelve questions, two minutes, nothing identifying. Answer honestly and this page hands you your shadow profile: what kind of operator you are and what to do next. The citywide count gets read out at Sharathon.

01Where does your AI use live?
02How often does AI touch your work?
03Does anyone at work know the full extent?
04What goes into your tools?
05Anything running without you watching?
06Honest hours saved in a normal week?
07Where does the saved time go?
08Has AI ever been confidently wrong in something that mattered?
09If you keep it quiet, why?
10If a sanctioned path were as good as yours, would you switch?
11Would you show your workflow to other builders if credit was guaranteed and punishment was off the table?
12Want in on building the sanctioned version for this city?
0 of 12 answered · anonymous, always
Part four · The amnesty

The promises.

Companies run amnesties with 30-day windows and executive memos. A city needs a standing monthly gathering with promises kept in public. These are ours.

Nothing shown gets punished

Bring the workflow that's been sitting in the background. Nobody here is your boss, and nobody cares whose account it runs on. The point is what you built.

Credit travels with the workflow

Show it and it's yours forever. Your name stays on it when someone else runs it. We don't claw back.

No silent ratchet

Revealing that something takes you thirty minutes will never become a demand that everything take thirty minutes. Saved time belongs to the person who saved it.

Free at the door

Lights On sessions cost nothing. The amnesty is the front door, and businesses pay for delivery when they want the rails installed.

Part five · Where the lights come on

Bring the workflow that's been chillin' in the background.

Sharathon is SD AI Studio's free monthly show-and-tell. Builders of every level showing what they made this month. Seeing each other's workflows is powerful for perspective and inspiration. Bring yours.

Sharathon #2 happened Sunday August 23 in La Mesa · 13 builders demoed · next date coming

Get the date, the recap, and the census read-out

events, builds, what's happening. no spam.