LLM Harms and Risks

What are the LLM Harms and Risks? Is there a good/reputable "list"? Is it ranked?

• list below is currently in my belief of descending-odds (the few top items are basically "100%, unless bubble pops and cost of usage kills usage")
• Probably need a 2-D space: size-of-harm vs odds-of-harm
• What we really need is an argumentation/claim-refactoring service.

(This looks like a decent list/definition of harms, need to review. And here's a list-of-lists.)

Even-worse customer service for everything (unaccountability machine)

Explosion of disinformation; increased odds of polarization, civil war.

Enshittification of all scalable search engines.

New form of cheating for people who want schooling credential rather than learning.

• "what appears to be a crisis is often the end of an illusion"; not-thinking tool

Abuse of privacy by goverment, insurance companies, data brokers.... Ouch

Data center electricity usage prolonging fossil fuel use, accelerating climate change

Electricity usage crowding-out: see (2025-11-21) Hunt World War AI

Bubble-pop crashing the economy/markets

Massive unemployment (%) of creatives (or is it post-ZIRP?)

Massive unemployment of developers (or is it post-ZIRP?)

• eliminating entry-level jobs, eating the seed-corn

Massive unemployment of middle-managers (or is it post-ZIRP?)

Existential risk from super-AI (super-empowered) controlled by authoritarian or mad-man (nuclear weapon, bioweapon, "crash the world economy", etc.)

Existential risk from AGI/ASI/Singularity/SkyNet with agency: p(doom)

Water usage, esp clean water: debunked? Still not great: (2026-03-05) AI Data Centers: What To Know About Their Water And Energy Use