Leaked Anthropic Model Presents 'Unprecedented Cybersecurity Risks'

https://gizmodo.com/leaked-anthropic-model-presents-unprecedented-cybersecurity-risks-much-to-pentagons-pleasure-2000739088

Comments

matanakapoMar 28, 2026, 3:54 PM
These hype cycles are exhausting
thankschuckMar 28, 2026, 3:51 PM
This is completely bullshit in my own opinion.

Either this is true and Anthropic is absolutely not valuable in keeping information secrets (so how about their users?!), or they are doing this to leverage fake hype (again...) about their IPO soon.

I am getting so tired of this fake hype...

xvectorMar 28, 2026, 10:53 PM
Why do you think this is hype when they were right about 4.5 and 4.6? SWEs using 4.6 barely ever write code anymore at my big tech.
thankschuckMar 28, 2026, 11:05 PM
Because of all the hype and bullshit we excessively get from 3 years now, and from all « AI companies ». I do not trust any « AI company » today, especially not when they are close of a big deal with their IPO.

Cool for them I guess, but big tech does not mean quality today (Meta, Apple, Google, Microsoft, Netflix… all those companies are doing s** today).

I continue to debug the code of my colleagues who use Claude everyday however. It is exhausting, but it saves the life of my low tech company today.

xvectorMar 29, 2026, 2:06 AM
I guess I'm not seeing the "hype and bullshit." Here is the Opus 4.6 press release. Which claims specifically read as "hype and bullshit" to you? https://www.anthropic.com/news/claude-opus-4-6
FinnucaneMar 28, 2026, 3:56 PM
"The fight has been about the Pentagon wanting to use Anthropic’s models even more than the company wants, which doesn’t seem like the thing you’d fight for if you thought the models presented security risks."

Well, remember who's running the show there.