Personal Generative AI Policy
This policy is a work in progress and I might change it at any time. It was last updated .
TL;DR: I don't use AI for any content on this site. I might use it otherwise but only with respect for others' time and my own cognitive health.
Motivation
Because LLMs, as well as many other forms of generative AI, are not inherently bound to truth, and instead to likelihood of output, it cannot be trusted on output alone. Many are rightfully concerned at the prospect that what they are reading, hearing, or seeing is the output of AI, as it represents a breach of trust by offloading verification from creator to viewer.
There are also many other ethical concerns surrounding AI in 2026, which are less relevant for this personal policy, but are worth mentioning for context:
- Most Gen-AI has been trained on copyrighted works such that it recreates it without attribution or permission, AKA plagiarism
- Gen-AI most common use is to replace human labor with something that doesn't say no, and/or dilute human output
- AI has been funded at massive externalized costs
- Datacenters used to serve AI have more negative than positive effects on the local community and environment, and therefore often have to be installed through undemocratic means
- LLMs have been trained for sycophancy, which leads to it agreeing with wrong premises in ways that make the user feel much more correct than they are, AKA "AI Psychosis"
Despite this, AI has its uses which, barring the ethical issues listed above, can make some new things possible or reduce tedium.
Generative AI Usage on this Site
I do not use generative AI for any content on this website.
The purpose of this website is to be representative of who I am and my thoughts. Usage of generative AI, and in particular LLMs, directly contradicts this purpose. Additionally, readers should be able to expect that whoever wrote the content has put in more effort than them in ensuring authenticity, quality and correctness, which AI also contradicts.
Generative AI Usage in Communications
I will never email/message someone with LLM output as though it's my own, UNLESS the recipient has evidently done so themselves. Direct messages imply interest in the response being from the recipient, so an LLM response disrespects the choice of medium. This implication does not follow for others' who do not follow this rule, therefore their communication channels are not bound by it.
In cases where LLM output is relevant, I tag it as AI generated with a short explanation for what I got out of it that can sufficiently describe the content on its own.