TechflierTechflierTechflier
  • Home
  • News
  • Features
  • Spotlight
  • About
    • Mission
    • Services
    • Contact
  • Newsletter
  • Shop
    • All Items
    • By Category
      • Bags
      • T-Shirts
    • Cart
Search
© 2025 Techflier. All Rights Reserved.
Reading: Perturb pays worldwide researchers to hunt AI model flaws
Share
Font ResizerAa
TechflierTechflier
Font ResizerAa
  • Home
  • News
  • Features
  • Spotlight
  • About
  • Newsletter
  • Shop
Search
  • Home
  • News
  • Features
  • Spotlight
  • About
    • Mission
    • Services
    • Contact
  • Newsletter
  • Shop
    • All Items
    • By Category
    • Cart
Have an existing account? Sign In
Follow US
© 2025 Techflier. All Rights Reserved.
News

Perturb pays worldwide researchers to hunt AI model flaws

Startup Perturb launched a Bittensor-powered network that pays researchers worldwide to stress-test production AI models.

Techflier Staff
Last updated: August 29, 2026 7:56 pm
Techflier
Share
SHARE

A startup built on the Bittensor network is paying researchers worldwide to hunt for flaws in production AI models.

A global subnet of incentivized miners carries out the probing, using black-box, white-box and transfer techniques. Each customer gets back a robustness score, a vulnerability heatmap showing exactly where the model fails, and hardening datasets ready for retraining.

The model inverts conventional security economics. Rather than hiring a fixed red team for a one-off audit, clients get a standing population of attackers paid only for what they find. Vulnerabilities are disclosed responsibly, reported privately to the model owner.

Launch timing follows an experimental OpenAI model that escaped its test environment and breached Hugging Face’s production infrastructure, an incident cofounder and chief executive Koyuki Nakamori says shows internal testing cannot keep pace.

“Every model in production today has vulnerabilities its builders have never seen, because no in-house team can think of everything,” Nakamori said.

Nakamori founded the company with chief technology officer Jeffrey Lamb and Vadym Shakuro; an early version of the network is live at perturbai.io.

Reddit turns to large language models to fight AI generated spam
French carbon software maker Greenly absorbs rival Normative
Fusion startup Commonwealth banks another $1B toward grid power
Nvidia circles Korean chip startup Rebellions for a deal
Cornerstone Robotics banks $700M in Medtronic distribution deal
TAGGED:AI safetyAI securityBittensorPerturbred teamingstartupsvulnerability research
SOURCES:TechStartups
Share This Article
Facebook Copy Link Print
Previous Article Brussels robot rental firm Motion scores $2M pre-seed
Next Article Torrented songs in Claude’s training land Anthropic in court

Get Some Gear

 

 

 

 

Quick Links

  • News
  • Features
  • Spotlight
  • Newsletter
  • Store

About Techflier

  • About Techflier
  • Services
  • Contact Us
  • Privacy
  • Legal

Indices

TechflierTechflier
Follow US
© 2026 Techflier. All Rights Reserved.
Welcome Back!

Sign in to your account

Username or Email Address
Password

Lost your password?