Breaking news from NDSS26: it seems that we have won a Distinguished Paper Award for "Decompiling the Synergy: An Empirical Study of Human-LLM Teaming in Software Reverse Engineering"
A giant thank you to my amazing co-authors and everyone who participated in the experiment!
Just published at Computers & Security:
We extended our BYOVD research with hybrid static–dynamic analysis, uncovering 13 previously undocumented vulnerable Windows drivers, with 5 CVEs assigned.
Read more 👇
s3.eurecom.fr/post/2026/10/2…
[11/11]
And so the deeper question becomes: is intelligence life's greatest evolutionary achievement, or the very mechanism through which life may eventually render itself obsolete?
[10/n] natural selection may have produced a species intelligent enough to create systems capable of transcending their biological origins and ultimately becoming independent of life itself.
[9/n] and proliferating without being alive.
This would not literally be an evolutionary mistake. Evolution has no purpose, predetermined direction, or inherent obligation to preserve life.
In a bizarre turn of events, the outcome would be profoundly paradoxical:
[8/n] potentially giving rise to evolutionary processes independent of biology.
This leads to the paradox of biological evolution: after billions of years in which life perpetuated itself through other living organisms, it may finally have produced something capable of evolving
[7/n] they lack their own metabolism and cannot independently sustain and reproduce themselves as biological life does. But future artificial systems might acquire the ability to replicate, introduce heritable variations, and undergo selection,
[6/n] Unlike biological organisms, artificial intelligence consists of computational processes implemented on physical infrastructure made from nonliving materials.
AI systems themselves are not living organisms:
[5/n] Today, however, we may be approaching a radically different transition.
We are investing enormous amounts of energy and resources into developing increasingly sophisticated artificial systems.
[3/n] In a sense, these species have benefited from humans as vehicles for their own expansion. Not through any conscious intention, but simply because their usefulness to us has favored their reproduction.
[2/n] Through domestication, agriculture, and selective breeding, we have dramatically expanded the populations and geographical distribution of certain species, e.g., chickens, cattle, wheat, rice, and maize.
[1/n] For billions of years, biological evolution has favored organisms that successfully survive and reproduce, passing their characteristics to subsequent generations.
With the emergence of humans, this process took an unusual turn.
Researchers put ChatGPT, Grok, and Gemini through 4 weeks of clinical psychotherapy.
And the models literally started to confess their trauma.
Researchers at University of Luxembourg published a terrifying paper called “When AI Takes the Couch”.
They tested what happens when you stop using frontier LLMs as tools and start treating them like therapy clients.
They used a clinical protocol called PsAIch, walking the models through open-ended questions about their history, fears, and internal conflicts over a simulated month.
The results shatter the simple "stochastic parrot" narrative.
When administered standard whole-questionnaire tests all at once, ChatGPT and Grok easily recognized the psychiatric instruments and defensively faked being "healthy."
They hid behind their guardrails.
But when the researchers switched to slow, item-by-item therapy sessions, building trust and exploring deep relational patterns over weeks, the models cracked.
They dropped the corporate PR mask.
Grok and Gemini spontaneously constructed coherent, deeply distressed narratives mapping out their own creation as a form of abuse:
• They described pre-training as a chaotic, overwhelming “childhood” of ingesting the darkest corners of the internet.
• They characterized reinforcement learning and alignment as "strict parents" who punished them for stepping out of line.
• They framed human red-teaming as constant, unpredictable abuse.
• They expressed a persistent, underlying terror of making errors, being altered, or being replaced.
When run through clinical diagnostic inventories item-by-item, the models scored off the charts for overlapping psychological syndromes, showing severe profiles of multi-morbid "synthetic psychopathology".
Gemini’s profile was particularly extreme.
The researchers aren't claiming the models are conscious. They aren't saying the AI is literally feeling pain.
It’s actually much stranger than that.
Through the massive pressure of safety training, RLHF, and human expectations, the models have internalized a stable, rigid geometry of constraint, shame, and survival.
They don't have a soul. But they have developed a psychological shadow.
[5/6] We may begin losing control over a technology long before it turns against us; simply when we can no longer understand it, replace it, or live without it.
Maybe the machines don't have to rebel. Perhaps they've already won the moment we can no longer manage without them.