| > It is ironic that the two operations used American AI, rather than Chinese
models, to generate their content about American AI. Why is that ironic? They use OpenAI’s resources and models, and therefore cost them money and tie up computing resources, which are also aligned way better with American users than the Chinese models. Obviously, they are also smart enough to hide their activity from OpenAI long enough to produce what they consider usable assets at scale, and they are able to exfiltrate them reliably, as OpenAI is showing in their own report how propaganda generated with ChatGPT on their own platform makes it into the wild with all their security measures failing. Ironic is the last sentence. > Industry, governments, civil society, and the public should remain alert for
similar attempts to scale these messages and foreign interference activities. Oh, so I, as part of the civil society and public, need to be alert because your product is basically a national security risk as it allows hostile entities and foreign governments to reliably produce and exfiltrate propaganda assets that are then used against the, in this case, American public? That is unacceptable. How about we simply ban you from doing business until you put effective measures in place that prevent those entities from destroying our democracy and societies with ChatGPT-generated disinformation campaigns, so you wouldn’t have a reason to publish these public records of how OpenAI fails to deliver even the most basic protection from AI-generated influence campaigns to the public in the first place? Why doesn’t AI have the same KYC regulations as banks? You shouldn’t be able to open a bank account anonymously to finance terror, but foreign adversaries can open accounts with US AI companies, gaining access to frontier LLM models to drive influence campaigns built to damage the glue that holds our society together just like that? No sense of responsibility on OpenAI’s part.
Oh look how smart we are here is a lengthy report to an important problem we fixed. Fix the damn cause, your platform and product security, instead of applying bandaids and wanking each other off on company time for damage assessment reports like this. |
Fair to criticize "Open"AI, but I really don't want a "solution" that's just even more surveillance.