Gary McGrawGo to the #MLsec movies Saturday
YouTube
The government does not know how to do this. NOBODY KNOWS HOW. #MLsec
"Meta is the only major U.S. developer of A.I. technology that has not reached an agreement to voluntarily share its models with the federal government for review. OpenAI, Anthropic, Google, xAI and Microsoft all agreed to submit their models to the government’s A.I. safety group, known as the Center for A.I. Standards and Innovation."
See https://berryvilleiml.com/results/no-security-meter-ai.pdf
This is hilarious. Counterfactual #MLsec versus #swsec failure reads like a blueprint of what will happen.
https://nesbitt.io/2026/06/26/incident-report-cve-2026-lgtm.html
The saga continues. Both the US government and Anthropic look like fools. #MLsec #swsec
https://www.nytimes.com/2026/06/26/technology/anthropic-mythos-government-restrictions.html
Hey now, beigification. #MLsec
“AI is always waiting for us to throw the stick for it to go and chase,” he continues. “It can only pull from everything that has thus far existed, and that photocopy of a photocopy will eventually start to be really dull and have no bearing on the human experience. So you can’t do it without us.”
https://www.theguardian.com/music/2026/jun/26/interpol-interview-elon-musk-fatherhood-ai-album
‘Elon Musk is dangerous and crazy. And I kind of used to like him’: Interpol on their political awakening – and making their masterpiece
They were a big 00s buzz band – but looked in danger of fading out. Empowered by fatherhood and anger at war and AI, the New Yorkers explain why they ‘really showed up’ again
@koehntopp oh so many reasons. Here are 81
Um...
I would not learn to #MLsec from the frontier people.
https://thehackernews.com/2026/06/openai-expands-daybreak-with-gpt-55.html
What is Whitebox #MLsec? How can we learn what #AI is doing inside the model?
https://berryvilleiml.com/2026/06/01/silver-bullet-security-podcast-157-tim-schulz/
BIML via video. Today we talked about https://arxiv.org/pdf/2505.03335
"Learning from (a sufficiently formal, but still open) environment. This approach will be limited to formal domains. Constraints allow us to ignore fundamental aspects of Turing completeness. The authors need to study some more computer science."
Harness adaptivity/evolution is a thing. Chinese work is good. #MLsec
As our research group reads and discusses scientific papers in #MLsec we add an entry to this (Very long) bibliography. We also actively curate a “top 5” list. Try searching this page with hashtags such as #TOP PAPER or #Recursive Pollution. (Last edit June 24, 2026 : 354 entries.)
