Tethics & Chill [online]: Yellow-Teaming 101—Threat Modeling for Externalities w/ Arm's Zach Lasiuk

​I’ve long admired economist Ronald Coase’s no-BS approach to externalities in his Theory of the Firm. In economics, an externality happens when a transaction impacts a third party who isn’t part of the deal—like pollution, privacy erosion, or, on the positive side of externalities, the smell of fresh bread from the bakery downstairs.

​Coase argued that markets can resolve these unintended consequences without regulation—if property rights are clearly defined and the transaction costs of bargaining are low.

​And you know what has a lot of externalities? Tech.

​The unintended consequences of technology—dark patterns, misinfo, bias, privacy trade-offs—often go unpriced, left for others to deal with after deployment. But expecting regulators to fix this feels...optimistic. Without waiting for Washington to wake up, there must be something that we, as technologists, can do to improve our own work.

​What would Coase do? What would it look like for tech to better govern itself? How can we, as the parties involved in the transaction and who know so much about it, enumerate our own externalities, price them into product design, and treat them as transaction costs—just as Coase might suggest?

​I call this idea Coasian Computing—or, put another way: threat modeling for externalities.

​Cybersecurity professionals do this all the time. They anticipate unintended consequences, hunt for anti-patterns, and build defenses before vulnerabilities turn into breaches. Why can’t engineers and product managers do the same? What if they proactively measured externalities, weighed them against business priorities like strategy and budget, and made eyes-wide-open decisions instead of externalizing costs onto society?

​Zach Lasiuk, a product manager at Arm (and a Tethics & Chill Austin OG!) took these questions and ran with them. What began as some rudimentary whiteboarding in my kitchen to map out a Strategy + Budget + Externality = Go/No-Go decision framework soon evolved into an externalities ChatBot in the early days of GPT and has now become a fully-fledged Yellow Teaming effort at one of the world’s leading chipmakers.

​So, how do we price in the costs of doing tech business before the bill lands on society? Let’s dig into the work Zach is doing, and maybe he’ll even run us through a scenario or two.

​Can't wait for this one! 🥳

​-Anastasia