OpenAI's June 2026 Report on Malicious Uses of AI [pdf]

Being very bullish on AI, I am not quite sure why this section is in this report. I have read decent amount of articles/news reports from reputed sources on real impact of AI data centers on electricity prices. For eg. https://youtu.be/iVGTGpKpykM?is=Hr2s386Hxn4yThlt

> The accounts we banned sought to influence two groups of audiences. They primarily targeted US audiences and generated English-language short comments and images claiming that data centers and AI applications were increasing electricity demand and causing higher costs for ordinary Americans…

Im glad OpenAI is publishing this sort of thing. This report is fascinating to read because it is so transparent - showing literal queries, images generated, etc. In the never ending, multifaceted tug of war (red/blue/anti-ai/anti-openai/anti-nuclear etc.) we have somehow lost sight of obvious facts like "its bad for foreign adversaries to operate influence campaigns." It's good for OpenAI to be transparent about this. It doesnt need a crazy conspiracy theory to explain it.

I think Americans in particular need to realize the discussion they are reading online is conducted by people that largely do not have America's best interests in mind - whether they just dont have a stake or they are actively working against it. This is not 2008 when Reddit was 80% Americans.

My point is not that we should all rally as a tribe or something. My point is Americans should stop assuming people engaging in discussion online are hoping America prospers. I single out the US because its a large part of the English speaking internet and used to be a majority group where this assumption was pretty fair. Online discussion spaces have become a lot more diverse over the past decade.

> 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.