Sriram Krishnan's exit: what shifting US AI policy means for SL devs
Sriram Krishnan is leaving his White House AI advisor role to start a new institution. Here's why a policy reshuffle in Washington still lands on your laptop in Sri Lanka.
The news that Sriram Krishnan is leaving his role as White House AI advisor sounds like inside-baseball Washington gossip, and from a desk in Colombo it's easy to scroll past. I don't think you should. When the person shaping a government's posture on AI changes seats, the rules that decide which models you can download, which APIs accept your card, and which export controls touch your GPU access tend to move with them.
According to TechCrunch, Krishnan is reportedly starting a new institution to keep shaping the Trump administration's AI policy from outside government. That detail is the interesting part, and the rest of this post is about why.
🔍 Why a Washington reshuffle reaches your laptop
US AI policy is not a domestic-only document. The frontier model labs, the cloud providers, and the chip designers you actually depend on are mostly American, so the rules written there become the defaults the rest of us inherit.
Here's the chain, concretely:
| Policy lever in the US | What it changes for you in Sri Lanka |
|---|---|
| Export controls on GPUs | Which cloud regions and instance types you can rent |
| Open-weight model stance | Whether you can legally download and self-host a model |
| API access / KYC rules | Whether your card and country are accepted at signup |
| Safety / licensing rules | How fast new model versions ship to everyone |
Key takeaway: You don't get a vote in US AI policy, but you absolutely live with its output. Watching who writes it is cheaper than being surprised by it.
When an advisor leaves to build a separate institution, the direction usually doesn't stop. It often gets more concentrated, because a dedicated outside body can push a single agenda harder than one person inside a busy administration.
🛠️ The free-tier reader's real exposure
I build on free and cheap tiers, and I suspect you do too. That's exactly the layer most sensitive to policy mood swings. Paid enterprise contracts have lawyers and SLAs. The free tier has terms-of-service that change on a Tuesday.
Three things I actually watch:
- Geographic availability. Plenty of model APIs still gate signups by country. A policy that tightens access rarely opens it for Sri Lanka first.
- Open weights vs closed. If an administration leans toward restricting open-weight releases, your ability to run a model locally, offline, on a borrowed GPU, shrinks.
- Compute cost. Export rules ripple into which data-center regions get the newest chips, and that ripples into the per-token price you eventually pay.
If a policy shift threatens your access to a hosted API, the hedge is not panic. It's having a local fallback you've already tested.
💡 The hedge: own your stack where you can
This is where a policy story turns into a to-do list. The cleanest insurance against someone else's regulation is reducing how many of your tools depend on a single foreign API that can revoke you.
I lean on this split:
| Layer | Cloud-API dependent | Runs on your device |
|---|---|---|
| Drafting text | Hosted LLM | Local model + your own editing |
| Transcription | Hosted speech API | Browser speech recognition |
| Image cleanup | Hosted SaaS | In-browser WASM tools |
| File conversion | Upload-based site | Client-side conversion |
A surprising amount of day-to-day work doesn't need a frontier model at all. The in-browser tools I keep at induwara.lk/tools run entirely on your machine, no signup, no foreign card, no policy that can switch them off. That's not a pitch, it's the point: the work that runs locally is the work no advisor in Washington can take away from you.
🌐 What a "new institution" usually signals
I won't pretend to know Krishnan's exact plans beyond what was reported, and I won't invent any. But the shape of the move is familiar. People who leave government to start a policy institution are usually betting that influence is more durable from the outside, where they can publish, convene, and lobby without the constraints of an official seat.
For builders, the practical reading is simple:
- The policy direction he favoured is unlikely to reverse just because the chair is empty.
- Expect continuity, possibly sharper continuity, not a clean reset.
- Watch what the new institution publishes once it exists. Those documents often preview the next round of rules before they're rules.
Bottom line: Treat think-tank output as an early-warning radar, not background noise. The white papers come before the regulations.
What this means for you
If you're a student, a freelancer, or a two-person team in Sri Lanka, you don't need to track every personnel change in Washington. You do need a posture, and the Sriram Krishnan departure is a good prompt to set one.
Do three small things this week:
- List your single points of failure. Which of your daily tools die if one US API closes its doors to your country? Write them down.
- Test one local replacement. Pick the riskiest dependency and run a self-hosted or in-browser alternative for a day. Knowing it works is the whole value.
- Bookmark the source, not the headline. Follow the institutions writing policy, not the outrage cycle reacting to it.
None of this requires a budget. It requires the habit of assuming that access granted by a foreign government's mood is access that can be withdrawn. Build a little resilience now, while the news is just a reshuffle and not a wall.