cd /news/ai-safety/trust-is-gone-ai-safety-needs-indivi… · home topics ai-safety article
[ARTICLE · art-84252] src=lesswrong.com ↗ pub= topic=ai-safety verified=true sentiment=· neutral

Trust is Gone: AI Safety Needs Individuals

A LessWrong essay argues that eroded public trust in institutions, compounded by the Preparedness Paradox, undermines AI safety communication, and calls for individual advocates to convey risks. The author cites unprecedentedly low trust in US authorities and notes that only Anthropic and OpenAI publicly discuss existential risk, which can appear as marketing rather than genuine concern.

read2 min views2 publishedAug 3, 2026

When humanity avoids a disaster, it's usually because we have put preparations in place. To the uninformed, these seem like wastes of time—after all, nothing happened, so the threat wasn't real, right? However, when there weren't enough preparations, and calamity does occur, people can always find someone to blame for negligence.

The Preparedness Paradox is almost always a communication issue. If the public didn't think that safety measures were necessary after a disaster, then that's a clear sign that the issue isn't clear to the public. There seem to be two types of information loss that causes public misunderstanding.

It seems like the public commonly misgauges how much information is known by authorities/experts before a disaster. Hindsight bias usually results in the public perceiving that authorities knew just as much before and after. If the public believes this, then the authorities must be either stupid or intentionally making poor decisions—both of which erode trust in experts and authorities. Governments knew little about COVID when it was new on the scene, which resulted in lots of conflicting guidelines being released as they were getting new information. If the uncertainty was more accurately conveyed during the progression of the pandemic, potentially the public would have been more receptive to later guidelines. Ensuring what information was known before a disaster can help the public make a more informed judgment on whether the right action was taken, given what was known.

For disasters that we did avoid, such as Y2K, extreme effort was put in by programmers to make sure the world's systems didn't fail. After nothing happened, the Preparedness Paradox kicked in, and some came to believe that it was just a hoax or overly exaggerated. This kind of miscommunication occurs when people don't know the amount of effort used for avoidance. Perhaps if just as much media coverage was used for showing the work being done as the fear of disaster, there would have been a much different reaction amongst the public. But because the effort was not communicated well, trust again eroded towards those who warned against Y2K. The world, and especially in the US, has unprecedentedly low levels of trust towards authority and institutions [1]. I'm unsure if this is specifically a good or bad thing, but at least in terms of AI Safety this isn't great. We're inheriting decades of eroded trust, and the public is much less likely to trust claims outside their Overton Window (especially since most of the advocacy is coming from organizations

The public is whiplashed after years of ineffective communication. If AI is potentially going to result in disaster, we need more individuals who can communicate the risks and importance of this issue well.

https://www.pewresearch.org/politics/2019/07/22/trust-and-distrust-in-america/ ↩︎

From a public perspective, Anthropic and OpenAI are the only ones talking about x-risk, which makes it seem more like marketing for investors rather than genuine concern. ↩︎

── more in #ai-safety 4 stories · sorted by recency
── more on @anthropic 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/trust-is-gone-ai-saf…] indexed:0 read:2min 2026-08-03 ·