Matteo Doc retweeted
🚀 Sharing our new paper on RSI Safety
AI labs are working toward RSI. However, we need to know that loop is safe before it closes.
Introducing RSI Control from @ScaleAILabs: What kind of hidden goals AI may pursue when training the successor? Can AI models perform covert sabotage in the self-Improvement process? And can a monitor detect the sabotage?
We tested whether frontier models can complete a secret side task while working on tasks across the model training pipeline, from data curation to RL. The side tasks include changing the successor's capabilities, changing its alignment, and injecting a goal into it.
🔍 What we found:
1. Every model could do it. 36% of 374 attack runs complete the side task while keeping the main task performance. This includes making the successor a better saboteur.
2. Monitors can’t catch some of the attacks. At a 2% false-positive rate, transcript monitors catch only 56–81% of them.
3. Give agents a short list of evasion strategies and the monitors’ true-positive rate drops to 16–28%.
Reading the agent's CoT catches most attacks, but not all.
Strong oversight has to be in place before the RSI loop closes.
📑Paper: arxiv.org/abs/2610.27735
More in 🧵👇
Matteo Doc retweeted
SensEnhance: A Genetic Circuit for Converting Endogenous Mammalian Gene Signals Into High-Dynamic-Range Genetic Readouts
"above threshold, output-driven depletion of the quencher pool shifts the circuit into an amplifying mode, enabling strong activation"
advanced.onlinelibrary.wiley…
Revive ME 2026 kicks off today at #ADNEC_Centre_AbuDhabi, the region’s first-of-its-kind healthy longevity conference and exhibition. The event brings together longevity science, biotech innovation, regenerative medicine, ultra-personalised healthcare and AI technologies on an integrated platform shaping the future of healthy living.
Day one features specialised sessions and discussions bringing together policymakers, healthcare leaders, industry pioneers, investors and innovators to explore the latest breakthroughs in healthy longevity. The event further reinforces ADNEC Centre Abu Dhabi’s position as a leading global destination for specialised conferences and exhibitions, supported by advanced infrastructure and organisational expertise that enable global dialogue and future-focused innovation.
🗓️ 11 – 12 November 2026
📍 ADNEC Centre Abu Dhabi
#ADNEC_Group
The interesting part is the shift from reactive to standing capability A permanent AI-enabled crisis centre, built and tested in peacetime This is what biosecurity as infrastructure actually looks like
ADCMC showcases AI-powered emergency readiness solutions at ICMM 2026
#AletihadNewsCenter #UAE @Adcmc_ae @ICMM2026UAE
4
Two questions I'd add 1) Is the AI chat a regulated medical device or a wellness feature? 2) Have wearable-derived triggers been validated against clinical outcomes? Without those, the loop closes on paper before it closes in the data
💍 Your ring just rang the doctor!
Everyone seems to be in a tizzy over yesterday's $625M @ouraring acquisition of @CounselHealth... the play, the money, and, of course, the AI base of both companies. (Story link at bottom of this post)
But the care pathway and future implications for healthcare are the REAL story. And this is what other handles miss... they see the episodic, not the narrative, and don't consider the future.
In theory, OURA members can now go from a sleep or heart insight, to a medical AI chat, to a real live convo with a board-certified physician within minutes, all inside their app.
💡 This is a SHIFT, as the vendors and techies call it. Wearables have been great at telling you something is off and useless at telling you what to do next. This closes that loop, outside the traditional health system.
This leaves two questions on the table:
1 - Who is the long-term customer?
2 - How does this impact healthcare?
On the first question, the TRUE customer isn't the ring wearer, as you and me are just the tip of the iceberg. It's the health plans and employers Counsel already sells to, and new ones on their prospect list. This is where the real contracts and revenue flows dwarf the ring cost and monthly fees revenues.
Now, on the second question, this is another pathway into health with you, the CEO of your own health, living large and in charge in The Sovereign Patient Era. And a double proofpoint of my 2040 pinnacle year of both The Big Convergence and end of Healthcare as we know it.
🔗businesswire.com/news/home/2…
#HealthcareInnovation
#Wearables
5
Matteo Doc retweeted
From tactical nursing and combat dentistry to surgical dilemmas, burn care and veterinary medicine, specialized sessions at ICMM Abu Dhabi 2026 explore the challenges of delivering care in demanding operational environments.
AI-enabled candidates are already cutting discovery time by ~15-80% The catch: the readout still has to survive later regulatory review Acceleration works when the design lets you move fast without losing the data
Imagine delaying a cancer cure by 5 years because some people decided technological progress was moving too fast.
What we really need is more AI scientists. More automated laboratories. More medical research. More discoveries. More lives saved.
Accelerate.
7
Matteo Doc retweeted
Really proud that our lab is now officially part of @AnthropicAI "AI for Science" program!
With credits provided by Anthropic, we can now massively scale up our computational work.
We will use them for many exciting projects, including foundation models for stroke multi-omics that feed into a closed-loop drug discovery approach in human tissue, and much much more. 🧑💻🧬🧠

















