One thing you should stop doing on social media right now:
tiulkanov.substack.com/p/one-thing-...
@tiulkanov.fr
LL.M., EU AI Act Course Instructor, CEN/CENELEC AI standards contributor, AI Governance consultant. Drafted AI and data regulation and policy for two nations. Ex-Council of Europe, ex-Deloitte Legal.
One thing you should stop doing on social media right now:
tiulkanov.substack.com/p/one-thing-...
And if you want a 1st class seat in the AI carriage next week, you can reserve it here: www.linkedin.com/events/amase...
We no longer a Digital Omnibus on our hands, as the AI Act amendments were carved out of the original joint ("Omni-") proposal to fast-track them. What we have now is really a Digital Duobus :)
Just one of the nuggets you've missed if you haven't been at my webinar on general-purpose AI models last Friday. And I won't blame you - it's peak vacations season.
The webinar recording is now available - exclusively for my Patreon subscribers: www.patreon.com/posts/webina...
My EU AI Act course has been pre-approved by IAPP for 13 CPE credits towards continued professional education requirements for these IAPP certifications: AIGP, CIPP/A, CIPP/C, CIPP/CN, CIPP/E, CIPP/US, CIPM, CIPT. The 5th cohort starts in Sept. Details and registration: aia.tiulkanov.info
People have been asking me for a one-pager about my EU AI Act course, so here it is. The 5th cohort starts in September, and for this run I'm adding coverage of the new GPAI Model Code of Practice and AI-specific cybersecurity considerations. See the details and book here: aia.tiulkanov.info
For those who missed my latest webinar on the interplay between the AI Act and the GDPR, the recording is available on my Patreon:
www.patreon.com/posts/webina...
Before you jump on the "thoughtless adoption of AI agents" bandwagon. This time it is about mailbox access, but you should think through a similar lens when granting any resource access privileges.
EC spokesperson for Tech sovereignty on the AI Act: "There is no stop-the-clock, there is no grace period, there is no pause... We have deadlines established in a legal text.": www.linkedin.com/posts/thomas...
U.S. tech lobbyists hysterical after the Senate overturns the proposed federal moratorium on state-level protections against reckless and abusive AI company practices:
datainnovation.org/2025/07/loss...
π§ Me 2,5 years ago: ChatGPT should be the tech that you don't trust that much. Verify all outputs before using them.
πΆ Sam Altman, 6 days ago: "People have a very high degree of trust in ChatGPT, which is interesting, because AI hallucinates. It should be the tech that you donβt trust that much."
And, conversely, whenever you see yet another AI slop β in a post, in an email β send them a link to this post. Let them know how you feel.
You're making yourself look like a fool and you're also showing lack of respect to your interlocutors, to their intelligence and their valuable time. Stop this now, everyone had enough of this nonsense!
AI can't replace human judgement. If you are lazy with your AI use and think it doesn't show β think again. It does show, it is very obvious.
It hit a nerve with everyone in the chat. This isnβt about being anti-AI β itβs about being anti-mediocre, anti-noise, anti-fake-efficiency, anti-stupid. The novelty has worn off. The overuse is real. And the burnout? Also real.
"Then I go online and see local businesses using AI-generated images for their ads. Iβve hit my limit.
On the bright side β at least Iβm spending less time on social media now."
Today, my fellow colleague expressed her frustration in our small group chat:
"I honestly canβt review one more document written by ChatGPT, or respond to yet another client comment also obviously written by a chatbot."
More and more people are getting mad as hell at AI slop and can't take it anymore.
Happy to announce that I am now a @PECB certified ISO/IEC 42001 AI Management System Implementer. www.credly.com/badges/9f918...
We know that training models on generated content causes model collapse (see autophagy).
So ironic: the more people populate the Internet with the results of these models put to their intended use, the less useful the internet becomes for training the next generation models.
π’German court gives green light to Meta AI training based on legitimate interest
Details: www.linkedin.com/posts/tyulka...
Explainer: www.linkedin.com/posts/tyulka...
Dear all, what happened to your critical thinking and ability to fact-check? A paper tissue and a cocktail pick, not something else, are clearly visible: www.gettyimages.de/detail/nachr...
NIAID's Integrated Research Facility is one of the few federal facilities charged with studying Ebola. Tuesday afternoon, all of its work was put on indefinite pause by Robert F. Kennedy Jr.'s department.
AI Act: risk management, conformity assessment and FRIA chart
A brief practical guide on distinguishing the substance and objectives of (1) risk management, (2) conformity assessment and (3) fundamental rights impact assessment under the EU AI Act (with a hi-res downloadable version of the graphic as a bonus):
www.linkedin.com/pulse/differ...
The European Commission is "assessing" Google's use of AI summaries at the top of its search results under the Digital Markets Act & EU copyright law
euractiv.com/section/tech...
<I feel we will also see scrutiny of AI services drawing on search indexes, esp wrt to personal data
I see your point, and you may be right in this reasoning. But I would not go as far as to use "fascism" as an accurate descriptor here. Maybe it is just me.
Nice, but what does that have to do with fascism?