Daniel Thilo Schroeder

Pentagon used Anthropic's Claude in the operation to capture Maduro? The operation included bombing targets in Caracas. Anthropic: "we cannot comment on whether Claude was used for any specific operation." The "safe AI" company hit the battlefield faster than it could update its terms of service?

Cyborg propaganda =
Verified human accounts + AI-generated articulation + centralized coordination.

Not botnets.
Not traditional astroturfing.
A hybrid model that exploits the human-in-the-loop loophole.

We outline mechanisms, risks, and governance paths.

doi.org/10.31234/osf...

Experts warn of threat to democracy from ‘AI bot swarms’ infesting social media Misinformation technology could be deployed at scale to disrupt 2028 US presidential election, AI researchers say

www.theguardian.com/technology/2...

My paper on the risks of influence operations, disinformation, and propaganda accepted for publication in the Journal for Intelligence, Propaganda, and Security Studies.

Democracy doesn’t require perfect truth—but it does require something more fragile: independent voices. The “wisdom of crowds” depends on independence between judgments. If a single actor can speak through thousands of inauthentic accounts, the apparent consensus of the crowd stops being informative

We have a new paper in Science today on how malicious AI swarms can threaten democracy.

AI systems can already coordinate autonomously, infiltrate communities, and fabricate social consensus.

www.science.org/doi/10.1126/...

Led by @kunstjonas.bsky.social & @daniel-thilo.bsky.social!

After years monitoring Twitter/X, I’ve gathered ~150 TB of research data on public discourse (Covid-19, Ukraine, George Floyd, etc.). X’s legal limits prevent sharing it with the scientific community. If you know lawful ways or partnerships to help us to make it research accessible, please reach out

www.nature.com/immersive/d4...

Teaching LLMs to Plan: Logical Chain-of-Thought Instruction Tuning for Symbolic Planning Large language models (LLMs) have demonstrated impressive capabilities across diverse tasks, yet their ability to perform structured symbolic planning remains limited, particularly in domains requirin...

This one is interesting.

Apparently, there is still room for improvement when it comes to symbolic planning/reasoning. arxiv.org/abs/2509.13351

Neural cellular automata: applications to biology and beyond classical AI Neural Cellular Automata (NCA) represent a powerful framework for modeling biological self-organization, extending classical rule-based systems with trainable, differentiable (or evolvable) update rul...

I read about those a while ago, but I think it might make sense to invest some time again.

Neural cellular automata are societies of tiny agents that preserve and repair global structure by propagating information locally.

arxiv.org/abs/2509.11131

We examined the role of social media in the amplification of violent, divisive discourse during Ethiopia’s Tigray conflict.

www.tandfonline.com/doi/pdf/10.1...

And I proudly did not use any LLMs for this post.

YouTube video by California Institute for Machine Consciousness Salon with Stephen Wolfram × Joscha Bach_What if we had bigger brains.

youtu.be/-G1SdsRXL7k

This book looks great! Can't wait to read.

(and thrilled to be a part of it. @nourkteily.bsky.social & I have a chapter on dehumanization you might be interested in!)

#PrejudiceResearch #PsychSciSky

Congrats to the Editors on curating such a strong list of chapters.

New preprint: How Malicious AI Swarms Can Threaten Democracy 🧪
osf.io/preprints/os...

When AI Goes Rogue: From Blackmailing Bosses to Bot Swarms Remember that scene in 2001: A Space Odyssey where HAL refuses to open the pod bay doors? Well, turns out that was optimistic.

Our paper is featured in this article on rogue AI agents and bot swarms:
www.theneuron.ai/explainer-ar...

Discussion here: news.ycombinator.com/item?id=4433...

🚨In PNAS🚨
The right often accuses fact-checkers of political bias
But we analyzed Community Notes on Musk's X and found posts flagged as "misleading" are 2.3x more likely to be written by Reps than Dems!
The issue is Reps sharing misinformation, not fact-checker bias...
www.pnas.org/doi/10.1073/...

ALERTE : 19 chercheurs de très haut niveau alertent sur la menace que représentent pour les démocraties les essaims d’IA malveillantes (« Ai Swarms »), qui peuvent se coordonner, échapper à toute détection, créer des informations convaincantes, et manipuler les élections. 1/3

Pre-print by @daniel-thilo.bsky.social et al. argues that the truth can be bent more quickly than it was in the past. I rarely go that far out on a limb, but in this case it is no longer a question of whether these scenarios will materialise. This is the world we live in and we need to prepare now.👇

Thanks for sharing Stefan!

Thanks for sharing!

@baronca.bsky.social @nachristakis.bsky.social @dgarcia.eu arcia.eu @amit-goldenberg.bsky.social @yarakyrychenko.bsky.social @ninalutz.bsky.social @garymarcus.bsky.social @gordpennycook.bsky.social @dgrand.bsky.social @summerfieldlab.bsky.social @audreyt.org @profsanderlinden.bsky.social

The combination of #artificialintelligence and #socialmedia poses a threat to democracy.

Our new paper explains how AI swarms can fabricate grassroots consensus, fragment shared reality, engage in mass harassment, interfer with elections, and erode institutional trust: osf.io/preprints/os...

I was looking forward to this Dagstuhl Seminar for quite some time. There is something to sitting somewhere in the middle of nowhere for a week and having discussions with a pile of very smart people.

"OpenAI's o1 Playing Codenames"

AI is showing some skills in games like Codenames, but it struggles to really understand teamwork and strategy, which makes it less impressive.
Article | Discussion

YouTube video by Curt Jaimungal Why The "Godfather of AI" Now Fears His Own Creation | Geoffrey Hinton

I am a big Goeffrey Hinton believer and only hope he's not right

www.youtube.com/watch?v=b_DU...

I agree with this. @hossenfelder.bsky.social

Screenshot of a paper in the International Journal of Press/Politics. Title: When Do Parties Lie? Misinformation and Radical-Right Populism Across 26 Countries. Authors: Petter Törnberg and Juliana Chueri. Abstract: The spread of misinformation has emerged as a global concern. Academic attention has recently shifted to emphasize the role of political elites as drivers of misinformation. Yet, little is known of the relationship between party politics and the spread of misinformation—in part due to a dearth of cross-national empirical data needed for comparative study. This article examines which parties are more likely to spread misinformation, by drawing on a comprehensive database of 32M tweets from parliamentarians in 26 countries, spanning 6years and several election periods. The dataset is combined with external databases such as Parlgov and V-Dem, linking the spread of misinformation to detailed information about political parties and cabinets, thus enabling a comparative politics approach to misinformation. Using multilevel analysis with random country intercepts, we find that radical-right populism is the strongest determinant for the propensity to spread misinformation. Populism, left- wing populism, and right-wing politics are not linked to the spread of misinformation. These results suggest that political misinformation should be understood as part and parcel of the current wave of radical right populism, and its opposition to liberal democratic institution.

This multi-year, multi-country analysis shows that right-wing populist politicians spread more online misinfo than their mainstream counterparts and explains how “the rise of political misinformation is associated to the recent wave of radical-right populism.”

journals.sagepub.com/doi/full/10....

Something I add to my when-I-am-retired-and-live-in-a-summer-house-in-italy-reading-list reading list.

~250 pages about LLMs FOR FREE!

arxiv.org/pdf/2501.09223

Unsere Webseite zur Bundestagswahl ist online! Hier findet ihr unsere Telegram-Datenbank-Dashboards sowie laufende Analysen zu Desinformationskampagnen und rechtsextremen Einflussnahmen.