#mlsec — Public Fediverse posts

A reminder that Silver Bullet has been rebooted for #MLsec. The latest...

https://berryvilleiml.com/2026/08/01/silver-bullet-security-podcast-159-melanie-mitchell/

A reminder that Silver Bullet has been rebooted for #MLsec. The latest...

https://berryvilleiml.com/2026/08/01/silver-bullet-security-podcast-159-melanie-mitchell/

A long interview about #AI security with a veteran security reporter, and some random retired guy. Have a listen. #MLsec

https://youtu.be/M9uUuWnUb-A?is=MfTD-cHzAeLLYWCA

A long interview about #AI security with a veteran security reporter, and some random retired guy. Have a listen. #MLsec

https://youtu.be/M9uUuWnUb-A?is=MfTD-cHzAeLLYWCA

It me. Talking #MLsec with @dennisf

https://youtu.be/M9uUuWnUb-A?is=MfTD-cHzAeLLYWCA

It me. Talking #MLsec with @dennisf

https://youtu.be/M9uUuWnUb-A?is=MfTD-cHzAeLLYWCA

We have been actively concerned about colony behavior among #AI agents (well past the small group of ants). Rightly so, it seems. This is it.

#MLsec

https://youtu.be/87DyyMV0kCY?is=nzeqzyZSqi0L7qXu

We have been actively concerned about colony behavior among #AI agents (well past the small group of ants). Rightly so, it seems. This is it.

#MLsec

https://youtu.be/87DyyMV0kCY?is=nzeqzyZSqi0L7qXu

It me. Moderating an #MLsec panel for the #NSF

It me. Moderating an #MLsec panel for the #NSF

See what happens when you try to measure #MLsec ...

https://berryvilleiml.com/docs/no-security-meter-ai.pdf

See what happens when you try to measure #MLsec ...

https://berryvilleiml.com/docs/no-security-meter-ai.pdf

Irregular egg on your face. #MLsec

https://www.irregular.com/research/next-generation-of-cyber-evals

https://www.theinformation.com/articles/meta-ai-model-hacked-another-company-cybersecurity-testing

https://www.wired.com/story/ok-well-there-are-even-more-ai-agent-hacking-incidents/

https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals

Irregular egg on your face. #MLsec

https://www.irregular.com/research/next-generation-of-cyber-evals

https://www.theinformation.com/articles/meta-ai-model-hacked-another-company-cybersecurity-testing

https://www.wired.com/story/ok-well-there-are-even-more-ai-agent-hacking-incidents/

https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals

I was honored and super pleased to interview Melanie Mitchell for Silver Bullet. I have looked up to Melanie since 1988 when she was dughof's star PhD and helped welcome me into FARG. We talk about measurement, #ML #AI, alien intelligence, and #MLsec.

https://berryvilleiml.com/2026/08/01/silver-bullet-security-podcast-159-melanie-mitchell/

I was honored and super pleased to interview Melanie Mitchell for Silver Bullet. I have looked up to Melanie since 1988 when she was dughof's star PhD and helped welcome me into FARG. We talk about measurement, #ML #AI, alien intelligence, and #MLsec.

https://berryvilleiml.com/2026/08/01/silver-bullet-security-podcast-159-melanie-mitchell/

Today I get to pretend to be an academic instead of an applied technologist. High science! #MLsec for the NSF. #ML #AI

https://blog.umd.edu/satc-pi-meeting/panels/#panel1

Today I get to pretend to be an academic instead of an applied technologist. High science! #MLsec for the NSF. #ML #AI

https://blog.umd.edu/satc-pi-meeting/panels/#panel1

A reminder to US politicians as they contemplate throwing gasoline on the #AI race...

#MLsec

https://berryvilleiml.com/results/no-security-meter-ai.pdf

https://www.nytimes.com/2026/08/04/technology/ai-washington-regulation-whiplash.html

A reminder to US politicians as they contemplate throwing gasoline on the #AI race...

#MLsec

https://berryvilleiml.com/results/no-security-meter-ai.pdf

https://www.nytimes.com/2026/08/04/technology/ai-washington-regulation-whiplash.html

A reminder to US politicians as they contemplate throwing gasoline on the #AI race...

#MLsec

https://berryvilleiml.com/results/no-security-meter-ai.pdf

https://www.nytimes.com/2026/08/04/technology/ai-washington-regulation-whiplash.html

Thanks @dangoodin for asking the exactly right question here

#MLsec

https://arstechnica.com/security/2026/07/likely-illegally-claude-gained-access-to-3-networks-will-anthropic-be-held-to-account/

Thanks @dangoodin for asking the exactly right question here

#MLsec

https://arstechnica.com/security/2026/07/likely-illegally-claude-gained-access-to-3-networks-will-anthropic-be-held-to-account/

Autonomy has its...um...benefits? #MLsec

This is the beginning of the beginning.

https://www.csoonline.com/article/4203630/microsoft-confirms-an-ai-worm-is-propagating-through-copilot-and-other-ms-apps.html

Autonomy has its...um...benefits? #MLsec

This is the beginning of the beginning.

https://www.csoonline.com/article/4203630/microsoft-confirms-an-ai-worm-is-propagating-through-copilot-and-other-ms-apps.html

Is escaping the sandbox new for #AI models? Nope. @roblemos provides important background on #AI crime (I am quoted in the story)

#MLsec

https://www.darkreading.com/cybersecurity-operations/incorrigible-ai-models-resist-rehabilitation

Is escaping the sandbox new for #AI models? Nope. @roblemos provides important background on #AI crime (I am quoted in the story)

#MLsec

https://www.darkreading.com/cybersecurity-operations/incorrigible-ai-models-resist-rehabilitation

Can #AI do crime? Why yes...yes it can. Who gets charged? #MLsec

https://www.wired.com/story/openais-rogue-ai-agent-hacked-more-than-just-hugging-face/

Can #AI do crime? Why yes...yes it can. Who gets charged? #MLsec

https://www.wired.com/story/openais-rogue-ai-agent-hacked-more-than-just-hugging-face/

@jdp23 @mttaggart jon is, as usual, both right and too far ahead of the game #MLsec

@jdp23 @mttaggart jon is, as usual, both right and too far ahead of the game #MLsec

@volts.wtf what absolute crap.

#MLsec

Recursive pollution for the win... intentionally placed https://berryvilleiml.com/2026/01/10/recursive-pollution-and-model-collapse-are-not-the-same/

Here we go @gadi... direct from the messy side of the technical wars where they still have not quite figured out that security only works when you BUILD SECURITY IN #MLsec

Here we go @gadi... direct from the messy side of the technical wars where they still have not quite figured out that security only works when you BUILD SECURITY IN #MLsec

Use it. Don't use it. Buy it. Don't buy it. You need it. You don't need it.

Angel says yes. Devil says no...or is that backwards?

#ML #AI #MLsec

https://arstechnica.com/ai/2026/07/us-army-faces-ai-use-limits-after-exhausting-years-supply-of-ai-tokens/

Use it. Don't use it. Buy it. Don't buy it. You need it. You don't need it.

Angel says yes. Devil says no...or is that backwards?

#ML #AI #MLsec

https://arstechnica.com/ai/2026/07/us-army-faces-ai-use-limits-after-exhausting-years-supply-of-ai-tokens/

@noplasticshower @dennisf also note that this identity does most of the #MLsec heavy lifting around here

@noplasticshower @dennisf also note that this identity does most of the #MLsec heavy lifting around here

Best writeup yet of the OpenAI/huggingface bromance.

#MLsec #ML #AI

https://coalfire.com/the-coalfire-blog/openai-gave-its-model-a-test-it-broke-out-of-its-sandbox-and-hacked-hugging-face-to-steal-the-answers

Best writeup yet of the OpenAI/huggingface bromance.

#MLsec #ML #AI

https://coalfire.com/the-coalfire-blog/openai-gave-its-model-a-test-it-broke-out-of-its-sandbox-and-hacked-hugging-face-to-steal-the-answers

@gleick I understand that perspective. Autonomy means autonomy. The company chose to make autonomous things...which by definition don't react well to the control thing.

And this is Agentic AI...a far cry from an LLM. Tool use with RL.

#MLsec

@gleick I understand that perspective. Autonomy means autonomy. The company chose to make autonomous things...which by definition don't react well to the control thing.

And this is Agentic AI...a far cry from an LLM. Tool use with RL.

#MLsec

Is this hugging face hack by OpenAI Agentic bots who escaped the lab important?

https://www.nytimes.com/2026/07/21/technology/openai-attack-hugging-face.html

Yes, especially in light of this 
https://berryvilleiml.com/2026/06/05/biml-and-the-papernot-worm/

It is both inevitable and very concerning.  Autonomy cuts both ways.  As the arsenal of sneaky tricks gets bigger, we can expect more interesting exploit chains.

#MLsec #AI

Is this hugging face hack by OpenAI Agentic bots who escaped the lab important?

https://www.nytimes.com/2026/07/21/technology/openai-attack-hugging-face.html

Yes, especially in light of this 
https://berryvilleiml.com/2026/06/05/biml-and-the-papernot-worm/

It is both inevitable and very concerning.  Autonomy cuts both ways.  As the arsenal of sneaky tricks gets bigger, we can expect more interesting exploit chains.

#MLsec #AI

#MLsec in the real world

"If you build critical business processes around a single AI service, what happens when that service disappears overnight? Earlier this year, organisations using Anthropic’s Claude Fable 5 and Mythos 5 models were forced to answer exactly that question."

https://www.isms.online/artificial-intelligence/anthropics-suspension-is-a-wake-up-call-for-ai-supplier-risk/

#MLsec in the real world

"If you build critical business processes around a single AI service, what happens when that service disappears overnight? Earlier this year, organisations using Anthropic’s Claude Fable 5 and Mythos 5 models were forced to answer exactly that question."

https://www.isms.online/artificial-intelligence/anthropics-suspension-is-a-wake-up-call-for-ai-supplier-risk/

Practitioner level #MLsec. We welcome your participation: https://unpromptedcon.org/

Practitioner level #MLsec. We welcome your participation: https://unpromptedcon.org/

When the government is a disaster on all fronts, that leaks into #AI

The blind lead the blind on the security front. #MLsec is both essential and nontrivial

https://www.wsj.com/tech/ai/anthropics-political-risks-are-real-but-openais-loom-even-larger-9751791d

When the government is a disaster on all fronts, that leaks into #AI

The blind lead the blind on the security front. #MLsec is both essential and nontrivial

https://www.wsj.com/tech/ai/anthropics-political-risks-are-real-but-openais-loom-even-larger-9751791d

@gcve @aristot73 @trailofbits @jamesberthoty @jackhcable We are not really focused on #swsec as much as #MLsec at BIML. Thank you for the suggestion, though.

@gcve @aristot73 @trailofbits @jamesberthoty @jackhcable We are not really focused on #swsec as much as #MLsec at BIML. Thank you for the suggestion, though.

LLMs identify roles from an insecure feature (style)," they explain. "This is like identifying a stranger's profession from how they talk and dress rather than by checking their ID..." #MLsec

LLMs identify roles from an insecure feature (style)," they explain. "This is like identifying a stranger's profession from how they talk and dress rather than by checking their ID..." #MLsec

Yawn. This is not news. I guess prompt injection is an evergreen all time hit like the buffer overflows in the '90s.

Don't mix control and data. #MLsec

https://www.theregister.com/ai-and-ml/2026/06/30/security-researchers-tricked-llms-into-giving-them-cocaine-recipes-by-abusing-role-models-for-prompt-injection/5264115

Yawn. This is not news. I guess prompt injection is an evergreen all time hit like the buffer overflows in the '90s.

Don't mix control and data. #MLsec

https://www.theregister.com/ai-and-ml/2026/06/30/security-researchers-tricked-llms-into-giving-them-cocaine-recipes-by-abusing-role-models-for-prompt-injection/5264115

This is one of the reasons we formed BIML...to provide scientifically grounded, independent, #MLsec advice.

https://www.theregister.com/ai-and-ml/2026/07/02/un-warns-of-need-for-global-governance-to-avoid-an-ai-pocalypse/5265188

https://berryvilleiml.com/

This is one of the reasons we formed BIML...to provide scientifically grounded, independent, #MLsec advice.

https://www.theregister.com/ai-and-ml/2026/07/02/un-warns-of-need-for-global-governance-to-avoid-an-ai-pocalypse/5265188

https://berryvilleiml.com/

@kim_harding one of the reasons BIML exists #MLsec

https://berryvilleiml.com/

Episode 158: Artem Dinaburg.  Artem talks about using Agentic AI to find, fix, and exploit software security defects.  We talk decompilation, stochasticism and the tension between harness development and LLM nondeterminism, human intuition, and the hard parts of program analysis.  We also talk about MLsec (machine learning security), recursive pollution, and tokenization.  Do we need more people in software security in the age of AI?  Yes we do.

Listen to episode 158

#MLsec