Everyone’s sleeping on one of Bittensor’s best use cases
Everyone talks about Bittensor through the same handful of buzzwords: intelligence, agents, decentralized training, inference, GPUs. Fine, I get it — those categories are exciting and easy to explain in a tweet.
But there’s a use case I think most people are completely sleeping on: cybersecurity.
Think about it. Cybersecurity isn’t a problem you solve once and walk away from. It’s a permanent arms race — hackers keep innovating, defenders keep patching, hackers find a way around the patch, and the whole thing starts over again. Forever.
Wouldn’t it be great if you could turn that arms race into your advantage instead of your biggest cost center? That’s basically what Bittensor lets you do.
A subnet isn’t really an “AI product” the way people usually picture it — it’s a competition. Miners submit work, validators score it, and Yuma Consensus decides who gets paid based on how good that work actually is.
That’s exactly why cybersecurity fits Bittensor so well. Security is adversarial by nature — you don’t want one static model guarding the gate, you want dozens of independent attempts running around the clock, real pressure from every angle, creative attackers pushing against creative defenders, and a system that only pays for results.
That’s why I keep coming back to Bitsec, Yanez and RedTeam.
All three run on Yuma-accelerated subnets — RedTeam is subnet 61, Bitsec is subnet 60, Yanez MIID is subnet 54. Per Yuma’s own subnet page: RedTeam tackles real-world cybersecurity challenges, Bitsec handles AI-powered code vulnerability detection, and Yanez MIID generates synthetic identities to train and test anti-fraud and identity-recognition models.
Three very different jobs. Bitsec goes after code vulnerabilities. Yanez goes after identity, fraud, compliance and proof of humanhood. RedTeam goes after adversarial stuff — bot detection, VPN detection, device-level attack and defense.
Different surfaces, same underlying idea: take cybersecurity R&D and turn it into an open competition, paid for by incentives instead of payroll.
Why this fits Bittensor so well
Old-school security companies run into the same wall every time: R&D is slow and expensive because it’s all headcount. Want more vulnerability research? Hire more researchers. Want more red-teaming? Hire more red-teamers. Every bit of extra output means another round of hiring, salaries, management, coordination.
Bittensor just blows that model up.
Instead of hiring, you define the task, set up the incentive, and let miners show up from anywhere in the world to compete for it — while validators grade the work. You don’t need to know in advance who’s going to crack it or how. You just build the competition so the best work rises to the top, because the money follows results, not resumes.
Wouldn’t it be great if you didn’t have to guess who your best researcher was going to be, and could instead just let a global crowd fight it out and pay whoever actually delivers? That’s the whole idea, and it matters even more in security specifically — so much of the useful work here is exploratory by nature. You don’t know where the bug is until someone finds it. You don’t know which bypass works until someone tries it. You don’t know which identity system breaks under which attack until it actually breaks.
I wrote about this exact idea in an earlier piece, “The Real Superpower of Bittensor” — decentralization by itself isn’t really the advantage. The real trick is splitting the innovation pipeline into two separate machines: a competitive global exploration engine, and a focused integration engine that turns the winners into something usable.
Cybersecurity might be the cleanest example of that idea I’ve seen yet.
Bitsec: AI-powered vulnerability discovery
Of the three, Bitsec’s job is the simplest to explain: find vulnerabilities in code, then fix them.
It’s built as an ecosystem for AI-powered code vulnerability detection — right now focused on subnet codebases and smart contracts, with plans to branch out into other codebases and vulnerability types down the line.
And honestly? This is exactly the kind of thing Bittensor should be great at.
Why? Because finding vulnerabilities rewards having a bunch of different approaches running at once. One miner runs static analysis. Another fine-tunes a model for it. Another builds an agent framework. Another mixes fuzzing with symbolic analysis and some manual reasoning on top. Bitsec doesn’t force one method — miners can throw whatever works at the problem: machine learning, fine-tuned models, agent frameworks, static analysis, all of it.
Compare that to a normal centralized security tool, stuck with one company’s model and one methodology. A subnet turns the whole thing into an open arena where a dozen different methods are all hunting the same bugs at the same time.
And recent news makes this way more than a nice theory.
Bitsec claims it found more vulnerabilities than a jailbroken instance of Claude Fable 5 in a head-to-head test. HackerNoon reported the test ran against a live paying client’s codebase — Fable 5 flagged around 60 issues, Bitsec found over 160, including five critical and ten high-severity bugs Fable 5 missed completely. Worth flagging the caveat too: this wasn’t Anthropic’s actual shipped product, it was a jailbroken version running outside its intended guardrails.
That caveat matters, so don’t go treating this like some clean academic benchmark.
But the bigger signal is real: Bitsec isn’t just claiming decentralized security could work someday, it’s running against actual codebases, producing actual findings, and going head-to-head with frontier-model vulnerability discovery.
Think about it. Mythos/Fable was used to discover a multi-year existing existential code vulnerability in the ZEC shielded pool. It basically allowed people to create out of thin air new ZEC coins without anyone knowing it. Not exactly what you want from a cryptocurrency claiming to be a store of value because of its scarcity…
Anthropic spent months warning everyone that their new model was powerful enough to be a genuine threat to digital infrastructure. So it’s honestly a little wild that a small startup, stitching together a handful of open-source AI agents, apparently beat that same flagship model at the exact thing Anthropic was warning about… What does it mean if open-source software plus Bittensor can out-compete closed-source models on cybersecurity? Big implications — but that’s a conversation for another article.
There’s a nice real-world proof point buried in here too: subnet-to-subnet usage. Yanez has publicly said it hired Bitsec to stress-test Yanez MIID’s incentive structure before mainnet launch. According to Yanez founder Jose Caldera, Bitsec confirmed vulnerabilities the team already knew about — and found new ones nobody had caught, some of which could have caused real problems at launch.
That’s the internal Bittensor economy actually working the way you’d want it to: one subnet making another subnet stronger!
Bitsec is still early-stage. There are signs of revenue coming in from audit contests, bug bounties and paid scanning work, but I wouldn’t put it in the same “confirmed revenue funding token buybacks” bucket as Yanez or RedTeam just yet — not without clearer reporting.
Strategically though? This might be one of the cleanest use cases Bittensor has: continuous, competitive, AI-assisted code security, running around the clock.
Yanez: identity, fraud and proof of humanhood
Yanez MIID doesn’t fit “cybersecurity” in the narrow sense — no firewall logs, no code audits. But it’s absolutely part of the modern security stack. Worth checking out yourself: yanez.ai.
It started out solving a financial crime problem: generating high-quality inorganic (fake but realistic) identities to test, tune and validate fraud detection, sanctions screening and broader compliance systems. Yanez’s own pitch is simple — financial crime prevention lives or dies on how well systems detect fraudulent identities, catch money laundering, and manage regulatory risk. And to actually test that properly, institutions need diverse, controlled datasets of fake identities to throw at their own systems.
I think this is a genuinely serious use case, because a lot of financial institutions are still testing their compliance systems against weak, static, incomplete data. Meanwhile attackers keep moving — fraud rings adapt, sanctions lists get updated, deepfakes keep getting better, and identity verification quietly falls behind attack patterns nobody planned for.
Yanez’s pitch, basically: generate the adversarial identity data yourself, and pressure-test your systems before actual attackers do it for you.
Their compliance product page claims their financial crime vulnerability scanner cuts testing and tuning costs by 90%, drops execution time from 1,000+ hours down to under 100, and produces dynamic multilingual datasets for testing and documentation.
More recently the story’s expanded past compliance testing into proof of humanhood, proof of uniqueness and agentic delegation. Yanez is building an app for cryptographically verifiable, hardware-agnostic, decentralized proof of humanhood — built on the assumption that autonomous agents are about to be everywhere. Zero-knowledge proofs, no biometric data leaving your device, decentralized uniqueness checks, delegation chains for AI agents acting on a human’s behalf.
That’s not a small pivot. It puts Yanez right in the middle of a question that’s only going to get bigger: how does a digital system know whether an action came from a real human, a fake identity, an AI model, or an authorized agent acting for someone else?
That question is going to matter everywhere — payments, airdrops, voting, KYC, DAO governance, financial approvals, agent-to-agent transactions, fraud prevention, all of it.
Yanez also landed two solid partnerships this year. In April 2026 they teamed up with BitMind to build a face deepfake detection model on Bittensor. In June 2026 they partnered with Nexartis to help organizations verify whether digital actions, content and transactions come from a real person, an AI model, or an authorized agent.
Those aren’t the kind of partnerships you get for being “an interesting crypto experiment” — they’re the kind you get for building actual commercial security infrastructure.
On revenue, Yanez looks like one of the more serious operations out there. Their July 2025 seed round brought in $900,000, earmarked for subnet 54 operations, R&D, marketing, and building a treasury of TAO plus the subnet’s own alpha tokens — money that also powers the Yanez Compliance platform.
Podcast summaries and Yanez’s own social posts describe real revenue flowing into alpha buybacks for the treasury, including a public claim of “20%+ of receivables in buybacks since end of March 2026.”
I’d still like to see a proper monthly dashboard — revenue, receivables, tokens bought, treasury balance, all in one place (though there is a public wallet you can go check yourself). But the direction is clear: Yanez is actually trying to tie real commercial revenue to the value of its subnet token.
RedTeam: turning hackers into your R&D department
I really think RedTeam is one of the most obvious use cases for Bittensor, period.
Think about it. You’re building cybersecurity software in a world where hackers continuously innovate new methods to bypass protective software. Wouldn’t it be great to use those same hackers to improve your product? This is exactly what RedTeam does! Through a continuously evolving competition across different challenges, Innerworks — the company behind RedTeam — manages to improve their product continually.
Innerworks provides cybersecurity software to, among others, private messaging apps. On X, they recently wrote that their product is already helping a well-known messaging app protect more than 100 million users. They claim plenty of clients and solid revenue, but I’d like to see actual proof of that. As of July 2026, there’s just not enough transparency yet around partnerships, enterprise contracts and revenue, in my opinion. Hopefully that improves soon.
Now back to what Innerworks actually does with the product.
Attackers never stop evolving their methods for getting around existing protection, so the software has to keep evolving too. That’s where the subnet comes in — it stress-tests Innerworks’ current protection product. If miners find a way around some piece of the security it’s supposed to provide, Innerworks can fix it immediately and ship something better. An ever-improving product!
Because of the sheer quality and speed of Bittensor miners, Innerworks has a real edge over competitors who don’t have access to a team of global, world-class hackers acting as miners. Those miners are basically part of Innerworks’ R&D department — except Innerworks doesn’t pay them. The Bittensor chain does.
This is exactly the kind of loop I think Bittensor investors should actually care about:
- Commercial security product generates revenue.
- Part of that revenue buys subnet alpha.
- Higher alpha value improves miner incentives.
- Better miner incentives attract better security work.
- Better outputs improve the product.
- The product generates more revenue.
That right there is the subnet flywheel in its purest form.
And the broader point doesn’t hinge on one client announcement. RedTeam matters because it shows exactly how Bittensor can turn adversarial talent into a persistent, ongoing security-improvement loop.
Why these subnets “need” Bittensor
Could Bitsec, Yanez or RedTeam exist as normal startups without Bittensor?
Yes, in a limited sense.
- Bitsec could be an AI security scanner.
- Yanez could be an identity and compliance testing company.
- RedTeam could be a bug bounty or bot-detection startup.
But they wouldn’t have the same structure. Not even close.
Without Bittensor, each team has to internalize way more of the R&D cost. They hire more researchers, pay more engineers, manage more experiments, manually source more talent. Their exploration capacity scales with payroll — full stop.
With Bittensor, the exploration layer gets externalized into a market instead.
That’s the crucial difference.
Bittensor gives these subnets:
- A global labor market of miners who can contribute models, code, agents, heuristics, datasets and attack strategies.
- A scoring layer where validators compare outputs and reward measurable performance.
- A token incentive layer where useful work gets ongoing participation in a live subnet economy, not a one-off bounty.
- A flywheel between product revenue and miner quality, especially once revenue starts funding token buys or treasury accumulation.
- A way to make adversarial iteration continuous, instead of an annual audit or quarterly test.
This is why I think cybersecurity might actually be a better Bittensor use case than a lot of the flashier AI categories. You don’t need to philosophically defend decentralization here — you can just ask a simple question: would you rather have one internal team looking for weaknesses, or a global market of independent agents and researchers trying to outperform each other every single day?
For security, that’s not even a close call.
Why investors may be underestimating them
The subnet market right now often rewards narrative over fundamentals. I get why — Bittensor’s young, and investors chase categories that sound enormous: agents, inference, training, synthetic data, GPU markets. Cybersecurity sounds less glamorous by comparison, even though the market is huge and the need is immediate.
Gartner forecasted worldwide end-user spending on information security to hit $213 billion in 2025, growing another 12.5% in 2026 to $240 billion.
At the same time, AI is actively reshaping the threat landscape. Frontier models help defenders, sure, but they also lower the cost of vulnerability discovery, fraud, impersonation and automated attack chains. Veriff’s 2026 Identity Fraud Report found a 300% jump in digitally altered or AI-generated media, with impersonation fraud now accounting for more than 85% of attacks.
That’s the macro backdrop.
Now compare that to what’s actually happening on Bittensor.
Bitsec is tackling AI-assisted code vulnerability discovery, and just showed the world that it’s not the almighty Mythos model winning this fight — it’s a combination of open-source agents outperforming it on critical vulnerabilities.
Yanez is tackling identity, deepfakes, compliance, and human/agent verification. Their proof-of-humanness app could genuinely become critical infrastructure in a world full of AI agents.
RedTeam is tackling adversarial detection, bot behavior, VPN flows, and a continuously evolving set of security challenges.
None of this is some abstract idea hunting for a market. These are direct responses to problems that already exist, right now.
And here’s the part that really bugs me: all three of these subnets are priced under 0.006 TAO as I write this in July 2026. For comparison, the highest-priced subnets sit somewhere between 0.05 and 0.08 TAO — a factor of 10 higher. Meanwhile there are dozens of subnets with zero revenue and barely an active product priced well above these three. That just feels wrong to me.
Why the disconnect? Simple — cybersecurity subnets are just harder to understand than “decentralized inference” or “AI agents.” The outputs are technical. The wins are often invisible. When a security system works, nothing happens. There’s no viral demo, no shiny chatbot, no image of a robot shaking hands with the future.
But invisible infrastructure can be enormously valuable.
Security usually only gets appreciated after it fails. Bittensor investors shouldn’t wait around for that failure to notice the subnets actually trying to prevent it.
Conclusion: security is an adversarial game, and Bittensor is built for adversarial competition
The strongest Bittensor use cases aren’t necessarily the ones with the loudest narratives — they’re the ones where decentralized incentives solve a genuinely real coordination problem.
Cybersecurity has exactly that problem.
The world needs continuous vulnerability discovery, continuous identity testing, continuous bot detection, continuous fraud adaptation, continuous adversarial pressure. Traditional companies try to meet that need with internal teams, point-in-time audits, and closed tools.
Bitsec, Yanez and RedTeam are pointing at a different model entirely:
- A codebase gets scanned by a market of competing AI security agents.
- An identity system gets stress-tested by synthetic adversarial data.
- A bot-detection system keeps improving because miners get rewarded for finding both the bypasses and the defenses.
- Commercial revenue flows into subnet tokens, strengthening miner incentives and potentially building a self-reinforcing loop.
That’s why these cybersecurity subnets matter.
They’re not just “security projects that happen to run on Bittensor.” They’re a clean example of why Bittensor exists in the first place: turn useful intelligence work into an open market, reward whoever actually contributes, and let competition improve the product faster than any closed team ever could.
If Bittensor succeeds, cybersecurity might just be one of the first places the world actually notices.