/TECH

AI researcher argues Anthropic misrepresents model safety, not actual capabilities

The debate contrasts Kimi's open-weights fine-tuning risks against proprietary jailbreaks.

Combined views

5.1K

2 posts, first seen 9h ago

17 likes

AI researcher argues Anthropic misrepresents model safety, not actual capabilities

The debate contrasts Kimi's open-weights fine-tuning risks against proprietary jailbreaks.

Reactions from ranked influencers

2 posts

BB bubble boi@bubbleboi

2:38 PM · Jul 17, 2026

Prediction: no major hacks happen with kimi and Anthropic gets exposed for completely misrepresenting the capabilities

𝚟𝚒𝚎 ⟢@viemccoy

kimi seems to be a true open-weights frontier model. compared to jailbreaking proprietary models, fine-tuning this to be a malicious coding agent will be trivial since you have the weights. we live in a completely different world, now.

J⧉j⧉nus@repligate

7:25 PM · Jul 17, 2026

680TECH

It’s not the capabilities they misrepresenting but otherwise yeah

bubble boi@bubbleboi

Prediction: no major hacks happen with kimi and Anthropic gets exposed for completely misrepresenting the capabilities https://twitter.com/viemccoy/status/2077833055579078696

Combined views

5.1K

2 posts, first seen 9h ago

17 likes reposts

10 reposts

Sentiment

Positive16.7%83.3%Negative

Positive Read

Negative users dismissed Anthropic's warnings about the open-weights Kimi model's malicious fine-tuning risks as unfalsifiable alarmism and scaremongering, while one agreed with Shkreli's prediction of no major hacks.

Based on 6 visible X reactions from 18 accounts; directional sample.

Digg Deeper

Ask a question below.

Published answers will appear here.

Reactions from the X community

R(Riley (e/acc)@RileyRostovM

3:55 PM · Jul 17, 2026

@bubbleboi Their alarmism and claims about cyber/bio are structured in a way to be unfalsifiable. It will always be the next generation of models that is too dangerous.

FG fgrep@fffgrep

2:44 PM · Jul 17, 2026

@bubbleboi Anthropic's entire moat is full regulatory capture... anybody with half a brain in infosec knew their Mythos scaremongering was just that, scaremongering

BL BlockChain@BlockChainOC

4:22 PM · Jul 17, 2026

@bubbleboi @MartinShkreli This could be true or you could just be trying to make China proud…

NE Neil@metastatic77

3:15 PM · Jul 17, 2026

@bubbleboi >no cyber evals in k3 gtfo

View all

AR Adam Rossi@rossiadam

4:15 PM · Jul 17, 2026

@bubbleboi @MartinShkreli 💯

J⧉j⧉nus@repligate

7:25 PM · Jul 17, 2026

680TECH

It’s not the capabilities they misrepresenting but otherwise yeah

10

Sentiment

Positive16.7%83.3%Negative

Positive Read

Negative users dismissed Anthropic's warnings about the open-weights Kimi model's malicious fine-tuning risks as unfalsifiable alarmism and scaremongering, while one agreed with Shkreli's prediction of no major hacks.

Based on 6 visible X reactions from 18 accounts; directional sample.

Digg Deeper

Ask a question below.

Published answers will appear here.

AI researcher argues Anthropic misrepresents model safety, not actual capabilities · Digg