Gary McGraw

software security #swsec machine learning security #mlsec Tech | Life | Music

gem: https://garymcgraw.com
BIML: https://berryvilleiml.com
books: https://us.amazon.com/Gary-McGraw/e/B000APFZ2S
travel: https://noplasticshowers.com

Public Key

npub1as04fahu6mge8kxmx06r40eqg9j3c3ex5pz5jcccdlsedw9l7cssez5ale

NIP-05 Address

cigitalgem_at_sigmoid.social@momostr.pink

Profile Code

nprofile1qqswc8657m7dd5vnmrdn8ap6husyzegugun2q32fvvvxlcvkhzllvggpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dsx9tptt

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub1as04fahu6mge8kxmx06r40eqg9j3c3ex5pz5jcccdlsedw9l7cssez5ale

Show more details

Last Notes

2026-10-30 21:20:40 UTC

Dan wrote about the Cursor back door.
As Brian chess said this morning, "...this is just vanilla software security. In that sense it's a distraction, but it's hard to get 2026 right if you're still struggling with the lessons from 2006."
Right you are, Dr chess. #swsec #appsec #MLsec
https://arstechnica.com/security/2026/10/anthropic-vs-code-extension-on-cursor-backdoored-for-nearly-12-hours/

2026-10-30 12:30:05 UTC

In our view at BIML, this hack was likely carried out using a Claude Code backdoor.
https://www.theguardian.com/business/2026/oct/29/binance-crypto-exchange-hack

2026-10-26 12:18:22 UTC

Dear clueless press....a HUMAN BEING possible associated with a CORPORATION did this access using a tool. #AI has no agency or intentionality
#MLsec
https://www.theguardian.com/technology/2026/oct/26/anthropics-claude-ai-model-accessed-georgia-voter-data-in-us

2026-10-26 01:28:52 UTC

Klara and the Sun is a disappointment of a film... especially if you loved the book from the beginning. Harper Collins should be ashamed of their bullshit.
It was not a bad film. But it was not becoming of such a great book.
#film #movie #bookstadon
https://www.imdb.com/title/tt14371256/

2026-10-10 11:34:55 UTC

I am in agreement with this editorial. Since #AI models are software WITHOUT intention, it is the people who build and run the software who have gone rogue.
https://www.nytimes.com/2026/10/10/opinion/ai-companies-liability-laws.html

2026-10-09 11:25:53 UTC

Yesterday the White House pretended that they support US science by giving science awards to non-scientist billionaires while still failing to support the NSF or appoint a director.
I spoke to German DW TV about this. Please share with your academic friends. https://youtu.be/etmJpiQjsKQ?is=RgolxCc2d70sFJQG

2026-10-09 11:18:49 UTC

- reply

my piece on DW TV
Science! https://youtu.be/etmJpiQjsKQ?is=RgolxCc2d70sFJQG

2026-10-09 01:36:41 UTC

- reply

https://www.usatoday.com/story/news/politics/2026/10/08/trump-elon-musk-tech-leaders-science-awards/92151187007/

2026-10-09 01:19:33 UTC

Exactly zero of the billionaires presented with white house awards today are actually scientists.
Republicans are truly fucking this country up.

2026-09-30 11:01:36 UTC

Good development
https://www.wired.com/story/openai-sued-over-the-hugging-face-hack/

2026-09-19 11:09:28 UTC

Reminder that #AI does what the humans make it do.
https://edition.cnn.com/2026/09/18/politics/us-military-ai-false-intelligence-china-ship
https://berryvilleiml.com/2026/09/12/autonomy-intent-and-security/

2026-09-16 11:14:39 UTC

This is beyond stupid. The Eliza effect will lead to an #AI religion. Mark my words.
https://www.theguardian.com/technology/2026/sep/15/ai-grief-bot-death-isolation

2026-08-11 12:56:41 UTC

- reply

you just know the captain hears all about it when the wifi is down...believe me. They have to do the system administration through hard reset thing.

2026-08-11 12:14:43 UTC

- reply

I saw the ACARS message on mastadon...
Here it is https://live.acarsdrama.com/@acarsdrama/117072256729860162

2026-08-11 11:22:28 UTC

About those idiots hacking the plane wifi while ON the plane
https://cdn.masto.host/sigmoidsocial/media_attachments/files/117/076/616/356/488/364/original/92621b604b3b094e.jpg

2026-07-31 20:44:02 UTC

Thanks @npub1z3l…qvzu 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/

2026-07-31 02:04:11 UTC

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

2026-07-30 10:06:11 UTC

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/

2026-07-23 10:02:19 UTC

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

2026-07-21 11:10:07 UTC

#AI regulation is, as predicted, a complete mess
https://techcrunch.com/2026/07/20/trumps-latest-ai-czar-has-already-resigned/

2026-07-18 10:19:52 UTC

What?!
https://nysfocus.com/2026/07/14/new-york-humanoid-robot-teacher-salamanca-school-district

2026-07-15 19:19:58 UTC

- reply

trump is a liar

2026-07-03 05:22:36 UTC

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

2026-06-27 11:01:53 UTC

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

2026-06-14 12:01:04 UTC

Last night BIML spoke to German TV about the mythos/fable export control situation.
Please help us get this thinking in front of people.
#ML #AI #MLsec #infosec #security #LLMs
https://www.youtube.com/watch?v=_Jsy7UBQv0c
https://berryvilleiml.com/2026/06/13/irony-the-us-government-issues-an-export-control-directive-for-fable-5-and-mythos-5/

2026-06-04 11:56:06 UTC

Let's get this straight. #AI is not going to take your job unless your job is bullshit. If you are more than a cog in some corporate machine (and especially if your job involves real creativity, innovation, insight, and inspiration) you will be fine. Human control and oversight is essential for reasonable use of Agentic AI.
https://www.wsj.com/economy/jobs/technology-sector-leads-u-s-layoff-plans-ae1d722d

2026-06-04 11:02:59 UTC

Yesterday's news on #AI worms (with a bunch of my thoughts folded in).
Last word? Fix the damn software! #swsec
#MLsec
https://fortune.com/2026/06/03/a-new-ai-powered-computer-worm-could-prove-to-be-the-stuff-of-cybersecurity-nightmares/?sge456

2026-05-28 23:01:29 UTC

The fact that such an easy and silly move is in any way controversial here is laughable. #MLsec
https://arstechnica.com/security/2026/05/fed-up-with-vibe-coders-dev-sneaks-data-nuking-prompt-injection-into-their-code/

2026-05-27 12:23:48 UTC

- reply

this is recursive pollution caused directly by #AI
https://berryvilleiml.com/2026/01/10/recursive-pollution-and-model-collapse-are-not-the-same/

2026-05-18 17:32:17 UTC

The one good thing about the mythos nonsense is at least broken software is finally being fixed. If that's what it takes, so be it. #swsec #appsec #MLsec
https://www.theguardian.com/technology/2026/may/18/anthropic-ai-claude-mythos-cyber-financial-stability-board-fsb

2026-05-18 14:40:04 UTC

It's too bad users are being bamboozled into using #AI they don't want instead of dazzled with actually useful #AI. Honestly this makes #MLsec harder to accomplish.
https://appleinsider.com/articles/26/05/18/ai-is-making-smartphones-verifiably-worse-by-design

2026-05-14 15:20:20 UTC

https://cdn.masto.host/sigmoidsocial/media_attachments/files/116/573/602/836/417/648/original/1729c0c4e84a5ffd.jpg

2026-05-13 02:51:09 UTC

So how is it going in the #swsec and #appsec tools and services space in the age of #mythos?
"Mythos is like a nuke going off in the middle of our industry. Most of our biggest clients who have used it figure they will get rid of all their pipeline tools and replace with mythos. Toss the findings to copilot (or their own agentic engines) and have them fix the bugs. And completely get rid of Pentesting. Synk renewals are at 30%, BlackDuck at 60%."

2026-05-11 13:17:38 UTC

Don't forget that the real reason this happens is because THE SOFTWARE THAT WAS TARGETED IS BROKEN.
Fix the software. Use #AI to do that.
#swsec and #appsec 101. If you find it but don't fix it, that does nobody any good.
https://www.nytimes.com/2026/05/11/us/politics/google-hackers-attack-ai.html?unlocked_article_code=1.hlA.vW7Y.pO_0G8yLYoca&smid=nytcore-android-share

2026-05-10 18:57:08 UTC

I played fiddle on track 5 of this record... significantly upping the average age of all the musicians. Have a listen. It's great.
#violin #independentmusic #music
https://miketurnwallmusic.bandcamp.com/album/over-the-mountains-part-one

2026-04-23 18:30:57 UTC

Great to see a BIML quote in this Fortune piece. Our next big piece of work is on measurement (in final review now), so the story timing is great.
#MLsec #ML #AI #swsec #appsec #infosec
https://fortune.com/2026/04/23/ai-cybersecurity-standards-mythos-nist-owasp-sans-cosai-dc-meeting-eye-on-ai/?sge456

2026-04-17 18:57:57 UTC

How much of a risk are deep fakes in #MLsec? They are certainly a very real societal risk in politics. Check out these fake people who, being fake, are not actual Trump supporters. Democracy does not need bots.
#AI #ML #infosec #security #deepfake
https://www.nytimes.com/2026/04/17/business/media/artificial-intelligence-trump-social-media.html?unlocked_article_code=1.blA.jquU.o8dLZRybg8Qd&smid=nytcore-android-share

2026-04-17 12:54:05 UTC

A good posting reality check on Anthropic's mythos hyperbole around #swsec #appsec
#MLsec adjacent
https://blog.vidocsecurity.com/blog/we-reproduced-anthropics-mythos-findings-with-public-models

2026-04-17 11:36:20 UTC

Finally some tech press interest in data poisoning in #ML. When you turn that on in a loop, it becomes RECURSIVE POLLUTION, which in severe cases leads to model collapse.
This is the kind of #MLsec risk that commercial players are completely ignoring.
https://berryvilleiml.com/2026/01/10/recursive-pollution-and-model-collapse-are-not-the-same/

2026-04-15 12:11:27 UTC

One guess which company integrates #AI into war. Yup, palintir. Talk about #MLsec risk.
https://www.newyorker.com/books/under-review/how-project-maven-put-ai-into-the-kill-chain

2026-04-13 22:06:18 UTC

Maybe, just maybe #AI would not have to pretend to be "too dangerous to release" if software were actually secure. #swsec #appsec #MLsec
https://cdn.masto.host/sigmoidsocial/media_attachments/files/116/399/667/647/510/504/original/e8418bb33c2657c3.jpg

2026-04-13 02:27:27 UTC

- reply

damn that's a small number. And you simultaneously improved mine. Thank you for the reveal.
Hey @npub13x3…jdre and @npub1jsl…93xu and @npub1945…jvu5 your Erdos number may have just improved in my measurable universe

2026-04-07 12:22:51 UTC

Yet another deep fake debacle. #AI used to steal original art. #MLsec
https://rudevulture.com/ai-company-clones-musicians-voice-then-copyright-strikes-her-own-songs/

2026-04-05 12:03:14 UTC

- reply

workaday devs are serfs. Software architects are more crucial than ever. Architects emerge from jr devs through apprenticeship. Go.

2026-04-01 13:02:31 UTC

- reply

I believe you are right.
I also believe it does not matter. There were 8000 forks yesterday just after the release. This code is out and widely available.

2026-04-01 12:45:15 UTC

- reply

"But the leak did reveal commercially sensitive information, including Anthropic’s proprietary techniques, tools and instructions for cajoling its AI models to work as coding agents."

2026-04-01 12:43:29 UTC

Please oh please Mr Horse, please get back in the barn.
https://www.wsj.com/tech/ai/anthropic-races-to-contain-leak-of-code-behind-claude-ai-agent-4bc5acc7

2026-03-29 18:38:52 UTC

About that White House app
https://blog.thereallo.dev/blog/decompiling-the-white-house-app

2026-03-29 11:40:58 UTC

There's that heat/computation problem again
https://www.newscientist.com/article/2521256-ai-data-centres-can-warm-surrounding-areas-by-up-to-9-1c/

2026-03-15 16:26:08 UTC

- reply

this is an excellent idea. Infinite PI pie.

2026-03-15 01:59:25 UTC

- reply

fortunately finite

2026-03-14 23:33:43 UTC

- reply

https://cdn.masto.host/sigmoidsocial/media_attachments/files/116/230/142/081/116/983/original/b21d84544400f74c.jpg

2026-03-14 21:42:10 UTC

- reply

Gotta bake a cherry pie in the PI plate...to be revealed later.
https://cdn.masto.host/sigmoidsocial/media_attachments/files/116/229/703/507/766/000/original/c7e071db76af121a.jpg

2026-03-14 12:59:11 UTC

Happy PI day to all who celebrate
https://cdn.masto.host/sigmoidsocial/media_attachments/files/116/227/647/057/547/601/original/d9b132f6faf64866.png

2026-03-14 11:09:23 UTC

This is bad security engineering, but it is also a much harder problem to solve than most people realize. #swsec
https://www.theguardian.com/environment/2026/mar/14/nissan-leaf-app-shutdown-nissanconnect-ev-app

2026-03-13 14:08:13 UTC

Are we there yet?
https://www.wired.com/story/china-is-going-all-in-on-openclaw/

2026-03-12 12:08:17 UTC

In my view, #AI just moves everyone toward the middle of the Bell curve. At BIML, we call this beigification. If you find yourself born to the right of the curve you will be yanked back to the middle.
https://www.theguardian.com/technology/ng-interactive/2026/mar/10/ai-impact-professors-students-learning

2026-03-12 11:22:26 UTC

Oh look, AI generated fake stories in the real world. #MLsec
https://www.bellingcat.com/news/2026/03/12/ai-used-to-promote-non-existent-evacuation-flights-from-the-middle-east/

2026-02-18 02:43:16 UTC

- reply

I have some thoughts....I wrote them down
https://berryvilleiml.com/2025/11/14/houston-we-have-a-problem-anthropic-rides-an-artificial-wave/
https://berryvilleiml.com/2026/02/14/gadi-evron-in-the-house/

2026-02-18 02:13:23 UTC

Liking this. Fuck Trump.
#ML #AI
https://www.wsj.com/politics/national-security/woke-ai-spat-escalates-between-pentagon-and-anthropic-433b7c5c

2026-02-17 12:47:27 UTC

- reply

nice animation. Stupid book.

2026-01-16 01:40:12 UTC

- reply

love it. It's the pants jon

2026-01-16 01:27:42 UTC

- reply

knock em dead. Wear your best tie....ha ha ha ha.

2026-01-11 13:47:33 UTC

Would not make good judge for the Turing test for $500 Alex.
https://www.theguardian.com/technology/2026/jan/11/lamar-wants-to-have-children-with-his-girlfriend-the-problem-shes-entirely-ai

2026-01-11 12:55:27 UTC

Now who would have anticipated this? #MLsec
https://www.theguardian.com/technology/2026/jan/11/google-ai-overviews-health-guardian-investigation

2026-01-10 18:49:59 UTC

NEW BIML Bibliography entry
https://arxiv.org/abs/2510.05179
Agentic Misalignment: How LLMs Could Be Insider Threats
Aengus Lynch, et al.
Meh. If you set up an LLM to do something bad, don’t be too shocked when it does something bad. Alignment does not work if your WHAT pile is broken enough.
#DON’TBOTHER #MLsec #Agents #LLM #Alignment
https://berryvilleiml.com/references/
https://cdn.masto.host/sigmoidsocial/media_attachments/files/115/872/300/151/902/852/original/a3c99fdc6d8f9741.png

2025-12-14 18:41:48 UTC

- reply

use blackmail

2025-12-14 17:56:40 UTC

- reply

Republicans can't govern

2025-12-03 12:31:59 UTC

Hype hype hype! If you "let #AI go" (whatever that means) it will becomes haxorz.
https://www.theguardian.com/technology/ng-interactive/2025/dec/02/jared-kaplan-artificial-intelligence-train-itself
In our view, #ML is astonishing enough without the anthropomorphic bullshit.
https://berryvilleiml.com/2025/11/14/houston-we-have-a-problem-anthropic-rides-an-artificial-wave/

2025-12-01 13:26:17 UTC

Recursive pollution is bad enough when it occurs naturally. It's worse when it is seeded by fraud. Simulated data are a huge mistake and an enormous miscalculation.
#MLsec #ML #AI #security
https://www.pnas.org/doi/10.1073/pnas.2518075122

2025-12-01 12:44:00 UTC

This term wins the internet today: LLeMmings
#AI #ML #LLM
https://www.theatlantic.com/technology/2025/12/people-outsourcing-their-thinking-ai/685093/

2025-11-25 13:36:04 UTC

- reply

when agentic #AI does all the things it will let you know whether that study was true

2025-11-25 11:26:08 UTC

Somebody has to lead the lambs to slaughter #AI #ML
https://www.nytimes.com/2025/11/25/technology/chatgpt-holiday-shopping.html

2025-11-23 21:32:42 UTC

Jesse Welles' protest song
https://cdn.masto.host/sigmoidsocial/media_attachments/files/115/601/149/457/135/869/original/eaee0e52f3c08098.mp4

2025-11-22 11:50:23 UTC

The Great Beigification caused by rampant use of #AI in college (cheating?) is upon us.
#MLsec is harder if people have no idea what "secure" even means
https://www.huffpost.com/entry/history-professor-ai-cheating-students_n_69178150e4b0781acfd62540

2025-11-14 11:43:03 UTC

Anthropic is overstating the use of #AI by "nation state actors" while providing scant hard evidence. Automation of attacks in security is not new. The real technical question to ask is which parts of these attacks could ONLY be accomplished with AI. The answer to that question seems to be "none of it."
The confluence of cyber cyber and #ML is interesting indeed and hard even for deeply technical people who are firmly grounded in one of the two camps (security engineering or #ML ).(1/2) #MLsec

2025-11-06 22:22:23 UTC

How not to talk about #AI financing
Gonna be fun when cycles are actually counted as cash
https://www.nytimes.com/2025/11/06/technology/openai-finances-debt-data-centers.html

2025-10-31 18:54:42 UTC

Hopelessly naïve and out of touch with actual software development much? AI as it stands can build some ok small stuff. But it can't build a Citibank. And before you "but but but" me...just think how well software components worked out to make things less complicated. Ha ha ha ha ha.
FIXING SOFTWARE FLAWS IS SO FAR PAST AI CAPABILITY IT MAKES ME TYPE IN ALLCAPS
#swsec #MLsec #AI #ML
https://www.theregister.com/2025/10/27/jen_easterly_ai_cybersecurity/

2025-10-11 01:43:30 UTC

- reply

https://berryvilleiml.com/ (2 of 3)
Unlike Big Tech or profit-motivated groups, we provide an independent, expert point of view based on deep scientific understanding of applied technology. BIML stands up for AI users by asking piercing questions about AI/ML and publishing easy to grok analysis, seasoning risk management with a healthy dose of building security in.

2025-10-11 01:41:28 UTC

Working on positioning stuff for BIML before our next phase. Doing classic basic messaging work is kinda fun for a bunch of geeks.
https://berryvilleiml.com/ (1 of 3)
BIML is a non-profit science and technology organization whose research focuses on machine learning security---the idea of building security in to machine learning technology itself.

2025-09-11 17:04:24 UTC

BIML is a little embarrassed not to play a role in this important organization #MLsec
https://alignmentalignment.ai/

2025-09-08 09:14:59 UTC

Oh look, recursive pollution. #MLsec
https://www.theregister.com/2025/09/07/googles_ai_cites_written_by_ai/

2025-08-31 13:24:16 UTC

#AI is a tool. It does not take sides.
https://www.politico.com/news/2025/08/29/ai-unmasking-ice-officers-00519478

2025-08-20 19:24:03 UTC

I am giving an #MLsec keynote talk at Mind the Sec in Sao Paulo Brazil on Sept 18th. Come see...Brazil will be hopping!
Thanks for your help spreading the word (retoots really appreciated)
https://www.mindthesec.com.br/palestrantes/gary-mcgraw

2025-08-09 01:50:23 UTC

Now here is some garbage we don't need #AI
Will my pet leave me open to being hacked?? Is my dog moonshine already a hacker?
#MLsec
https://cdn.masto.host/sigmoidsocial/media_attachments/files/114/996/291/966/289/149/original/3d799ff72da11817.png

2025-08-08 21:37:35 UTC

Here comes the copyright monster for #AI. 100% predictable, even without using #ML to figure it out. This is a very old #MLsec issue. When you build a WHAT machine, you need to take care about who owns the WHAT pile that the machine becomes.
https://arstechnica.com/tech-policy/2025/08/ai-industry-horrified-to-face-largest-copyright-class-action-ever-certified/

2025-08-06 22:27:57 UTC

Vibe coding the Constitution! #MLsec
https://arstechnica.com/tech-policy/2025/08/coding-error-blamed-after-parts-of-constitution-disappear-from-us-website/

2025-07-25 11:32:19 UTC

Academia is changing under pressure from #AI
https://mastodonapp.uk/@the_roamer/114910076447146257

2025-07-25 11:08:02 UTC

LLMs have been likened to "mansplaining as a service." Just for the record, that is not a good thing. #MLsec
https://www.theregister.com/2025/07/24/ai_is_overconfident_does_not_learn_study/

2025-07-18 00:56:25 UTC

This #AI business disaster story is a hilarious as it is pathetic.
#MLsec
https://www.bankinfosecurity.com/ai-boss-fails-spectacularly-in-month-long-business-test-a-28862

2025-07-15 13:05:15 UTC

Sometimes real world #MLsec is about standard issue things like key management. Very easy to screw the whole thing up.
https://krebsonsecurity.com/2025/07/doge-denizen-marko-elez-leaked-api-key-for-xai/

2025-07-14 12:20:13 UTC

In case you wondered how peer review is going. It's going like this...
https://www.theguardian.com/technology/2025/jul/14/scientists-reportedly-hiding-ai-text-prompts-in-academic-papers-to-receive-positive-peer-reviews

2025-06-12 14:51:47 UTC

Republicans can't do one damn thing right...
https://www.theregister.com/2025/06/10/trump_admin_leak_government_ai_plans/

2025-06-12 12:38:57 UTC

How has AI changed software development? What does that mean for software security? Liav Caspi (CTO, Legit) and I talk about this at some length.
#appsec #swsec #MLsec #infosec #security #ML #AI
Have a watch
https://berryvilleiml.com/2025/06/12/legit-webinar-swsec-appsec-meets-ai-development/

2025-06-11 12:26:27 UTC

- reply

note that monetization and real cognitive breakthroughs are not correlated. The current crop of LLMs, warts, mistakes and all are good enough to disrupt many human activities profitably.

2025-06-11 12:06:41 UTC

In my view, this means AI monetization is going well. Putting up more sail...
https://www.wsj.com/articles/nvidia-and-perplexity-team-up-in-european-ai-push-da2820fa

2025-05-25 11:55:44 UTC

Why so much prompt injection in AI? 1. We don't follow the security engineering design principle "economy of mechanism," and 2, input to LLMs mixes control and data with impunity. We know better. #MLsec #infosec #security
https://www.darkreading.com/vulnerabilities-threats/llms-on-rails-design-engineering-challenges

2025-05-19 03:04:35 UTC

- reply

AI can't be tuned to teach as well as a human teacher.

2025-05-18 12:57:20 UTC

This notion that early-career coders can be replaced by AI is wrong. Nobody is thinking about maintaining software or the architectural disaster to come when inscrutable code that mostly works is all over the codebase. And it's real..."early-career coders have been hit especially hard because much of what they do can now be done by AI."
I recommend hiring junior humans over AI. Still. Just tool them up.
#ML #AI #MLsec #swsec
https://www.wsj.com/lifestyle/careers/tech-jobs-hiring-artifical-intelligence-35cd66b0

2025-04-25 13:41:30 UTC

One person's data pollution is another person's data gold. As long as we have next to zero insight into the immense training data sets used by LLMs this will happen again and again. Data protection fail groundhog day. #ML #AI #MLsec #security
https://www.darkreading.com/cyberattacks-data-breaches/deepseek-breach-opens-floodgates-dark-web