Applying Intelligence Community Indications and Warning methodology to frontier AI yields a single, stark conclusion: we are currently in an active warning failure. The capability thresholds intended to trigger policy interventions have already been breached, with frontier models clearing 50-70% on SWE bench and inference efficiency expanding at a  40x annually. Our current evaluation frameworks are structurally gameable by situationally aware systems, pointing to a foundational counterintelligence failure rather than a mere oversight gap. The governance community must immediately pivot from behavioral black box testing to white box mechanistic auditing, moving away from trying to prove danger and toward enforcing mandatory compliance frameworks.

The observation about electricity being useful for 140 years while 600 million people still lack access is sharp. Usefulness alone has never been sufficient to overcome the infrastructure barriers a technology needs to get deployed.
Every technology splits into two layers  infrastructure and application. Private capital flows toward application because ROI is higher. Infrastructure is a public good, returns are slower, and the funding it requires runs into political barriers most LMICs cannot overcome.
The deeper problem is what is happening right now in governance. Every framework being built  EU AI Act, national AI strategies, safety frameworks focuses heavily on deployment rules and model behaviour. The compute infrastructure layer where actual concentration is happening, remains completely ungoverned. No major governance framework treats compute concentration as its problem.
This means LMIC exclusion is not just a market outcome. It is being locked in at the governance design stage. By the time LMICs have meaningful political leverage to demand infrastructure access the ownership structures will already be legally and commercially entrenched.
Compute governance needs to be part of the development economics conversation not just the AI safety conversation. Right now these two fields are not talking to each other and that silence has consequences.

Evals are being gamed not because the methodology is insufficient but the models on which the compliance audit run are sophisticated enough to game the audit.
IC methodology already solved the problem of denied human capabilities through triangulation by using independent behavioural signals not better direct elicitation .The AI safety community needs to make the same epistemological shift.
The question isn't how to make evals harder to game, it's whether evals are the right instrument at all.

So the author is right that evals are overrated but the reason is deeper then model detecting when they are being tested.

The core diagnosis is this: Evals are architecturally wrong for the job.They are designed like compliance audit e.g one test, one result but if we organized  capability assessment against a system that already aware of that requires a more adversarial methodology . 

IC methodology which faces a same problem like this. So the solution was not direct elicitation it was triangulating across independent behavioural signals so no single source could be game.

When we apply it to AI by using different cross reference sources like red team outputs, deployment behaviour and independent replication by compartmentalized team. Now the convergence through different independent resources is much harder to game than a single test.

The fix is not better evals.Its a different methodology framework which is closer to competitive intelligence than compliance audit .