How AI Leads to the End of Secrets

I remember the first time that I used ChatGPT Deep Research. I wrote a pretty decent prompt, put it in deep research mode, and the AI went off for 45 minutes to think and dig around on the Internet.

When it came back with a 35-page report, I knew the world had changed.

Now — let’s be clear. ChatGPT Deep Research is a bit like a junior research analyst right out of school. I would take everything that analyst said with a grain of salt. I’d fact check anything that didn’t seem right. I’d try to guard against confirmation bias and dig deeper on anything that felt too good to be true.

But there was no mistaking the reality: $200/month for ChatGPT Pro was buying me the equivalent work of a junior research analyst, who was remarkably gifted at connecting the dots and finding meaning from little data points.

Immediately the thought hit me — is AI leading us to the end of secrets?

Secrets by Obscurity

For decades, the business world has depended on obscurity to keep secrets. If you weren’t one of the Magnificent Seven companies with reporters crawling all over you, it was possible to build new products or develop new initiatives, and make extensive progress before announcing it to your customers and to the world.

When I was CEO at Riskalyze, I’d do somewhere between 30–40 interviews a year with reporters or podcasters. Of course, I’d never leak our unannounced strategy to build that new product or enter that new market. But undoubtedly, a smart reporter or podcaster would ask a question, and wanting to be consistent with my projected self in the future, I’d make a few remarks that expressed openness to expanding in that direction.

There was no easy way to piece together the little nuggets I’d dropped on podcast X, with the data points I’d mentioned in my interview with journalist Y, with the LinkedIn groups one of our product managers was joining, and the job openings we’d posted on our careers page.

Those little micro-data points used to be very difficult to assemble into strategic insights.

Not any more.

Secrets Aren’t Dead. But Their Half-Life Is.

AI hasn’t made it impossible for a company to keep secrets. But it has dramatically shortened how long most secrets stay secret.

Strategic roadmaps? Easy to spot via hiring patterns, public messaging, interview comments, and partner activity.

Algorithmic breakthroughs? Models can now reverse-engineer behavior, scan research exhaust, look at newly-released API docs, and piece together working theories that are “close enough.”

Your latest pricing test? Already scraped, modeled, and mimicked by someone who’s never even logged into your admin panel.

AI doesn’t need a leak. It just needs breadcrumbs — and businesses leave those behind constantly. Stray comments, a shift in social media post wording, code commits, API behavior, conference chatter, shipping data, and patent filings…it’s all discoverable, siftable, and analyzable now.

The Formula

So, if you’re working on a secret project, how do we think about its discoverability in this new era?

I think the formula is something like this: Discoverability = Exposure × Observability × Inference ÷ Countermeasures

• Exposure: How many surfaces are emitting signals? Think repos, contracts, public demos, hiring trails.
• Observability: How machine-readable and real-time are those signals?
• Inference Power: This is where AI shows up — connecting dots that humans would miss.
• Countermeasures: The technical, legal, and operational steps you take to reduce discoverability.

AI dramatically increases inference and observability. That means our job as leaders is to reduce exposure and strengthen countermeasures — especially when secrets need to hold long enough to deliver value.

What Secrets Hold Up?

Not all secrets are doomed, and here are some categories of secrets where tomorrow’s edge still lives.

• Air-gapped work: If the bytes never leave, the model never sees them. Simple.
• Private, non-replicable data: Exclusive data sets, sensor networks, or customer flywheels no one else can copy.
• Privileged compute or fabrication: Custom chips, secured firmware, or closed-loop manufacturing that stays dark to the outside world.
• Compartmentalization: If no single system or person knows the full picture, even a leak gives partial signals.
• Execution-embedded secrets: The supply chain trick, or the weird process nuance that doesn’t live in code.

If you want durable advantage, don’t rely on keeping an idea quiet. Rely on making it hard to replicate, even when people know it exists.

The New Playbook: Secrets in the Age of AI

I suspect we’re in for a huge amount of change in how companies begin to protect their competitive secrets in the AI era. Corporate environments are going to start looking a lot more like military or intelligence ones.

We’ll see a lot more “default private” architecture — no public logs, limited CI/CD mirroring, zero egress by default.

We’ll see compartmentalized teams and credentials — Steve Jobs was famous for setting this up at Apple. Nobody has more than they need, and the blast radius from leaks is limited.

We’ll see more companies red-teaming their footprint regularly — using AI agents to scan what an adversary could infer from their public data, and building internal OPSEC bots that automate detection of leak vectors in meetings, Slack threads, and third-party integrations.

We’ll see organizations engaged in creating noise and misdirection — posting generalized or even misleading job descriptions, shipping decoy branches, or obfuscating API behaviors.

And most importantly, organizations will begin designing execution for speed — working to ensure that even if the idea leaks, they’re already 50% to 75% of the way ahead of any competitors seeking to fast follow.

The Bottom Line

AI collapses the cost of connecting dots. Inference is practically free. Scraping is trivial. Pattern detection is instant. That means your secrets — especially the idea-level ones — have a half-life compared to the past.

But that doesn’t mean you’re out of weapons. It means your strategy shifts.

From keeping ideas hidden, to making them hard to copy.

From strategy secrecy, to speed of execution.

From hoping for silence, to assuming fast leaks and winning anyway.

If someone connects the dots, make sure they’re still months behind.

That’s the new game. Play to win.