Moondream AI co-founder Vikhyat K. jokes that corporate AI safety means embedding 20 engineers within the NSA
Story Overview
Vikhyat K.'s quip uses the reported presence of Anthropic engineers inside the NSA to highlight a gap between the company's public safety messaging and its work on a cybersecurity model cleared for limited offensive uses, especially with the firm having just filed confidentially for an IPO.
A handful of engineers sit inside NSA facilities to customize models
Multiple outlets describe roughly six forward-deployed Anthropic staff guiding usage of the Mythos Preview model, though it remains unclear whether they join active operations; access is capped at about 40 organizations total.
The safety pause call lands right after the IPO filing and a new funding round
Anthropic submitted its S-1 on June 1 and closed a $65 billion round at a $965 billion valuation in late May, then published the global slowdown proposal on June 4-5 without specifying timelines or enforcement steps.
Many users mocked Anthropic's NSA cyberattack claims amid its safety push as a classic larp and bizarre pivot from prior messaging.
Most Activity
@vikhyatk Hahahahahahahahah holy shit
we care about safety. that’s why we have 20 FDEs embedded in the NSA to help with offensive cyberattacks
@vikhyatk honestly time for companies to rename departments of cybersecurity to cyberwar
@vikhyatk well sure, you don't want children running around with scissors unchaperoned! what are you gonna do, not sell them scissors?
@vikhyatk the pivot from "safety" to "embedded with the NSA" is a wild vibe shift lmao
@vikhyatk thats the "embedded" part they want us to glance past
@vikhyatk Classic larp, these agencies operate way above this pay grade.
@vikhyatk the phrasing just gets more creative the deeper u go into that world