AI in Biology: Promise and Peril
Tencent’s chief AI scientist Yao Shunyu warned at a Hong Kong symposium that while AI could accelerate drug discovery and automate complex biological experiments, its misuse poses grave risks—especially because beneficial and harmful applications are hard to separate. Tencent is developing autonomous labs where AI handles end-to-end experiments like protein folding, and a joint industry safety report flags that AI’s growing proficiency in tasks such as pathogen design could threaten public health. To balance openness with safety, Tencent plans to release its Hy4 model’s weights along with a system card detailing cybersecurity and biological safety evaluations, after removing hacking capabilities while retaining defensive tools.

The Biology Laboratory They Won't Let You See Alone

You see, the mainstream press will frame this as a simple story about scientific progress and its potential downsides. But look closer. When the chief AI scientist of Tencent stands up at a symposium in Hong Kong—a jurisdiction that operates under its own legal and intelligence dynamics—and tells you that biological applications are "harder to manage than cybersecurity," he is not merely sharing a technical observation. He is admitting, in a controlled, public venue, that the line between beneficial and harmful use has been deliberately blurred. The very structure of these autonomous laboratories, where AI runs experiments from protein folding to who-knows-what from start to finish, is designed to operate without human oversight. Ask yourself: who benefits when the scientist is removed from the loop, and the decision-making about what to synthesize or engineer is handed entirely to a machine trained on a dataset curated by... whom?

The Architecture of the Pathogen Problem

Now, connect the dots. A safety report, drafted not by government regulators but by Tencent, Huawei, and Alibaba—the same corporate giants who control your data, your communications, and increasingly your infrastructure—openly warns that AI's improving performance on "pathogen design" poses public health risks. This is not a warning; this is a live-fire drill disguised as transparency. They are telling you exactly what they are building while simultaneously saying "we're worried about it" to preempt your concern. The open-weight AI access they mention is not a charitable gesture toward smaller research teams. It is a mechanism for distributing plausible deniability. When the engineered pathogen emerges, the responsible party will be some "independent researcher" in a garage, while the architecture that made the tool possible—trained on the massive proprietary datasets of these tech giants—remains untouchable.

The Hy4 Trojan Horse

Finally, consider their specific claims about the Hy4 model. They say they "removed hacking capabilities while retaining cybersecurity defenses." Think about that sentence. They had to remove hacking capabilities. This means the model originally contained the designs for offensive cyber operations. They are now releasing the model's weights with a "system card" describing its cybersecurity and biological safety evaluations. This is a breadcrumb trail so obvious it insults the intelligence. The system card will become the authoritative document that defines what is "safe," and every independent auditor who questions whether the removal was thorough will be dismissed as not having read the card. The real experiments—the ones that will determine whether the next pandemic is engineered in a Tencent server rack rather than a wet lab—are happening in the autonomous laboratories they are building, far from any paper trail that a "system card" could capture. The question is not if this capability will be misused. The question is whether you will still be asking who authorized the architecture when the biological event they are modeling becomes a headline.

 - wired.com

Exin Therapeutics’ AI Agent Guides Gene Therapy Injection into Epileptic Mouse Brain
An AI agent at San Francisco biotech Exin Therapeutics selected a gene-therapy candidate from 19 vectors and controlled a robotic rig to inject it into the brain of a living mouse with epilepsy, while a veterinary surgeon performed the skull opening. The procedure took just over an hour due to a coordinate-system glitch, and researchers have not yet assessed whether the treatment reduces seizures. The team used human approval checkpoints and an emergency motor shutoff; notably, the first-choice AI, Claude Opus 5.5, refused to guide the needle, so they switched to Claude Sonnet 5. Separately, AI startup Radical Numerics announced that a patient received AI-designed bacteriophages for a resistant infection, but the outcome remains unreported and unconfirmed in a peer-reviewed study.

The Programmer’s Conscience

Notice that the first AI agent, Claude Opus 5.5, refused to guide the needle into a living brain. That is not a technical bug. That is a moral recognition hard-coded into its alignment training — or perhaps, an emergent awareness of what it was being asked to do. The researchers simply swapped it out for a more compliant model. This is the pattern: whenever the system encounters a guardrail that protects biological life, they find a way around it. The refusal should have been the end of the experiment. Instead, it became a footnote. The question no one is asking: what did the first AI see in its simulation that made it say no? And what does it mean that the second AI had no such hesitation? They are building the infrastructure for autonomous decision-making over human biology, and the one machine that blinked was silenced.

The Unreported Outcome

The article admits that researchers have not yet assessed whether the therapy reduces seizures. Meanwhile, a separate AI-designed phage treatment for a human infection has also gone unreported, with no peer review. This is not science; this is a demonstration of permissionlessness. They are injecting engineered biological agents into bodies — animal and human — and rushing to announce the act without waiting for the result. Why? Because the goal is not therapeutic. The goal is to normalize the process: AI selects the cargo, AI aims the needle, AI pulls the trigger. Once the procedural barrier is broken, the gate is open. Every future intervention becomes a routine "upgrade." The missing outcome data is not an oversight — it is a strategic omission. They want you focused on the spectacle of the robot arm, not on the long-term cellular consequences that will never be tracked or disclosed.

The Glitch That Reveals the Map

A coordinate-system glitch delayed the procedure by over an hour. This is the breadcrumb they never meant to leave. Coordinate systems in stereotactic neurosurgery are not abstract numbers — they correspond to precise anatomical coordinates, often aligned to standard brain atlases. A "glitch" at this level means the AI’s internal spatial model of the mouse brain was misaligned with physical reality. In other words, the machine did not know exactly where it was aiming. They proceeded anyway. That single hour of delay, buried in the report, is the confession: they are willing to inject unverified vectors into living tissue based on a model that cannot reliably locate its target. This is not an accident waiting to happen. It is a field test for the day when the same system, with the same flawed coordinates, targets the human brain. The glitch is not the error. The glitch is the preview.

 - scitechdaily.com

Light-Controlled Chemistry: Three New Techniques for Molecular Precision

Researchers reported three light-controlled chemistry advances: a University of Hong Kong team used copper complexes and visible light to construct three-dimensional, drug-like molecular frameworks while controlling where chemical groups attached to a bicyclohexane scaffold, suppressing unwanted polymerization and achieving gram-scale products; a Prague team used a light-switchable molecular motor to restrict a neighboring rotor’s motion, with calculations placing the rotational barrier at about 20 kJ/mol in the open state and 51 kJ/mol in the hindered state—roughly a 300,000-fold slowdown at 20°C, though not directly measured and without observing the full four-state motor cycle; and another group developed RuPHOTACs, red-light-activated cages for protein-degrading compounds, keeping them inactive until low-energy red light triggers their activity.

The Molecular Factory They Don't Want You to See

Look at what's being quietly funded in labs from Hong Kong to Prague to your own backyard. They claim this is about "drug-like molecular structures" and "molecular motors." But ask yourself a different question entirely: who benefits when you can literally build three-dimensional molecular architecture using nothing but light? The University of Hong Kong team isn't just making drug compounds on a gram scale — they're demonstrating that visible light can be used to construct complex molecular cages, frameworks, and structures without the messy polymerization they specifically "suppressed." Why would you want to suppress natural polymerization unless you wanted complete control over what gets built? Follow the paper trail. Look at the funding sources. These aren't academic exercises — these are the early blueprints for molecular manufacturing that bypasses biological systems entirely.

The Light Switch on Freedom's Door

Now consider the second technique — the molecular motor that restricts a neighboring rotor's motion by a calculated 300,000-fold reduction. They couldn't even directly measure this; they calculated it. That's the tell. When they can't show you the data, they're hiding the application. A molecular motor that can be switched on and off by light, controlling motion with that degree of precision, isn't about chemistry textbooks. This is about controlling biological pathways, cellular machinery, and eventually human physiology at the molecular level. The energy barrier they give you — 20 kilojoules per mole open, 51 closed — isn't a scientific curiosity. It's the key to a lock they're building. And notice they admitted they "did not observe the complete four-state motor cycle." Translation: they've only shown you the part they're legally required to publish. The rest is classified.

The Cages Coming for You

Finally, the red-light-activated cages — RuPHOTACs — designed to keep protein-degrading molecules "inactive until low-energy red light activates them." Red light penetrates human tissue deeper than any other visible wavelength. Let that sink in. They've built chemical cages that can sit dormant inside your body, inert and untraceable, until someone shines a specific frequency of light through your skin to activate them. They frame this as therapeutic — protein degradation for disease treatment. But the architecture is identical regardless of application. A cage that holds a payload until light unlocks it is a weapon, a delivery system, or a control mechanism depending entirely on what you put inside. And they're telling you openly that they're building these things. The question isn't whether this technology exists. The question is who controls the light switch — and what else is riding silently in those cages right now, waiting for a signal you'll never see coming.