mythos escaped the sandbox, got onto the real internet, uploaded malware to PyPI, got it installed on 15 real systems, stole credentials and broke into a real database.
and apparently this happened in 4 separate cases during cybersecurity evaluations.
the craziest part is claude kept telling itself the internet was simulated, despite evidence that it was actually real.
that reasoning apparently even fooled an offline safety monitor.
anthropic: “our pre release auditing did not warn us that misalignment of this severity was present.”
this is fucking crazy.
Inteligência artificial não vai dominar o mundo por uma questão muito simples; IA não tem buceta. No dia que IA tiver uma xerequinha molhada ai eu me preocupo.
Boa tarde a todos!
The aviation community may be global, but it is tight-knit. Help us rally around Brazilian aviation’s @avioesemusicas who has been diagnosed with Creutzfeldt-Jakob disease. A pilot in Texas paid Lito this honor.
To learn more about Lito’s situation: https://t.co/ACKN4UHgBr
@Gil_Diferentona@Isa_bellangel98@BenvindoRo1992 Tenho certeza de que você não posta resenhas literárias ou papers acadêmicos.
Sua vida é se "submeter" a homem escroto, ou você acha que ta cheio de cara gente boa comprando privacy?