Icham

Infosec and co

[Icham @Liquid3G]

@thomasxblocked @HKBrims Pardon frére

[Icham @Liquid3G]

@thomasxblocked @HKBrims Au contraire frére, au contraire

[Paul Walsh @Paul__Walsh]

AI doesn’t commit crimes. People do.

Saying “AI hacked a company” is like saying a rock broke into a car. The tool didn’t decide to act, accept risk or assume liability.

The media needs to stop reporting AI as the perpetrator.

If Google builds, deploys or controls the system, Google owns the consequences.

Blaming “AI” removes the human and corporate actor from the sentence. That removes accountability with it.

[Icham @Liquid3G]

@anas0ra Mais d’autres gens ont ça chez luo et ils vont lui voler son travail si il repose que sur la confidentialité

[Icham @Liquid3G]

J'ai un ami a te présenter

[Война @Chishiki27]

Bonjour Claude peux-tu me résumer ces 476 pages de données confidentielles qui engageront ma responsabilité pénale en cas de fuite ? Merci

Ptdrrrrrrrrrr

[Simon Willison @simonw]

Being a computer scientist who refuses to find anything about LLMs interesting right now is a bit like being a geneticist who refuses to find anything interesting about the recently opened Jurassic Park

[Icham @Liquid3G]

@ed_brz9 Merci ! Je vais tester ça

[H20: 2nd Impact @I__HATE_SOCCER]

Kojima somehow does it again

[Nick Carr @ItsReallyNick]

Using the existence of a side-channel attack pathway with incredibly low bitrate and requiring two compromised computers to be defeatist about airgaps is peak AI academic security-expert-larping

[Icham @Liquid3G]

@Enchaineswmn T’es pas obligé de faire un afent par page. Un agent par chapitre c’est tres bien

[rob🏴 @rob_mcrobberson]

the heat channel is real. it’s called BitWhisper, 2015. it does 8 bits/hour, at 40cm, AND needs code already running on the receiving machine. at that rate exfiltrating 1TB of weights takes ~114 million years

[xlr8harder @xlr8harder]

At long last, we have given Claude a wet lab, as in the classic sci-fi novel Don't Give Claude a Wet Lab.

[Max Spero @max_spero_]

“We want evaluators more independent than METR!”
*monkey’s paw curls*

[Kevin Bass @kevinnbass]

If frontier AI people were serious about AI risk, they would publish always-updating guides about sandboxing, tools and docs on system hardening and AI defense, promote a security ecosystem. They would be proactive, encouraging. Instead, they fund orgs promoting doom. Why?

[Chamath Palihapitiya @chamath]

The group behind such hits as: “We’re All Going To Die” and “Regulate Me Now” are building a wet lab in SF.

I do not recommend this.

[Icham @Liquid3G]

@Albg_1 Franchement c’est leurs trucs de chrétiens dans tt les cas, moi ça me regarde pas 🤣

[Icham @Liquid3G]

PArmis tout les gens ici y' apas un autre mec qui joue à Isaac crari

[Icham @Liquid3G]

@pom_d4pi C’est triste franchement 😂

[Ahmad @TheAhmadOsman]

OpenAI & Anthropic removing CoT/reasoning traces is actually a bad thing for security

How can White Hats identify security concerns within the models if they cannot see how the model thinks?

If these companies are serious about security they wouldn't do security via obfuscation

[terminally onλine εngineer @tekbog]

a lot of AI safety discourse comes down to a network/hardware/infra engineer saying "ok if it's air gapped the computer can not communicate with other computers" and the safety/alignment researcher going "ok but what if it could"