Premium Yebisu Beer, Born in 1890, Tokyo

Prions And Viral Vectors Will Be Skynet's Weapon Of Choice

Note: this writing is to flag an urgent biosecurity concern - obviously, many details have intentionally been left out to not be a document of inspiration.

We face an emerging biosecurity threat that combines the lethality of prion diseases with the transmissibility of engineered viruses. This threat has received insufficient attention despite representing one of the most dangerous bioweapon scenarios imaginable: a highly transmissible viral vector carrying genes that induce fatal prion diseases in infected hosts.

 

What are prions

Prions are infectious agents composed entirely of misfolded proteins. Unlike the usual pathogens such as bacteria or viruses, they contain no DNA or RNA, yet possess the ability to convert normal cellular proteins into their own lethal configuration. This process creates a cascade of neurodegeneration that is invariably fatal.

While prions are infectious and have the ability to reproduce by converting healthy proteins into copies of themselves, they are generally not considered “alive”. Though whether something counts as being “alive” or not is a target of rigorous debate, we can apply some existing definitions to analyze.

When applied to the classic “textbook” definition (and yes, I am well aware on what constitutes a “textbook definition” is also debatable) of life:

One can consider prions to be a “corrupted” type of protein that 1) has been misfolded to a lower energy status than their “useful” counterparts, and 2) contains a unique ability to act as a catalyst to convert other normal proteins into a lower energy form, which has the same structure as itself.

What makes them uniquely dangerous is the way they operate. Instead of attacking the host like traditional pathogens, they corrupt the host’s own proteins - turning the host’s own cellular machinery against itself. This process is slow, undetectable for years / decades, and impossible to stop once started.

Prions have 2 methods of initial expression within a host:

  1. Exposure from an external source, from where the prions start replicating within the host
  2. Carriers of genes that induce prion formation, where hosts with certain genes are susceptible for a spontaneous prion formation, from where further replication starts

The genes that induce prion formation have been actively studied (although some exact mechanisms still remain unknown). Such genes are well-characterized and a part of publicly available data.

 

The threat: viral delivery of prion-inducing genes

Recent exponential advances in gene editing and viral engineering have created a new threat vector that combines the worst aspects of both biological hazards. Technologies to embed specific genes into a various set of viral vectors are becoming increasingly accessible. Combined with the fact that the genes behind prion formation are publicly available, this is a potential recipe for a massive disaster.

Unlike naturally occurring prions, where transmission is limited between hosts, a viral vector carrying prion-inducing genes could theoretically spread through populations like any respiratory virus. Viral transmission and embeddings of prion genes could widely occur in the population for years without detection. By the time the first prion symptoms start showing up, a vast amount of population may already be housing an incurable, 100% fatal condition - basically a death sentence that spreads.

This is no risk to ignore - the technological basics for such already exist today:

Unlike legitimate research facilities that conform to replication-deficient vectors, not all entities may make the same choices.

For now, we may remain a bit assured since the steps do present significant research challenges (obviously we will not go into details on what the challenges are). However, this does not mean no action is needed. The theory is sound, and technologies are exponentially improving. All while we remain completely unaware of this specific construction, which may have civilization-wide impact.

 

The AGI amplification factor

The intersection of this construct with artificial general intelligence (AGI) development further multiplies the risk. An advanced - and heavily misaligned - AGI system would likely identify prion-viral vectors as its optimal arsenal, and possess capabilities that further amplify the threat.

From an optimization perspective, a prion-virus represent the most efficient solution to what a misaligned AGI might consider the “human problem”:

An advanced AI system would possess capabilities that make this threat orders of magnitude more dangerous than human-developed versions. AGIs with access to vast amounts of research data would instantly process all publicly available research on viral vectors and prion biology, identifying optimal combinations that might take human researchers years to discover. Unlike humans constrained by lab limitations and human error, an AGI could computationally model and optimize viral vectors and prion genes while minimizing physical lab footprints that may lead to detection.

 

This can be civilization-ending

The prion-virus represents a perfect combination of dangerous characteristics that potentially make it the biggest threat to humanity - more so than nuclear war, climate change, vanilla AI takeover, etc.. When combined with AGI development, the threat to humanity becomes insurmountable.

 

Delayed recognition until too late & detection challenges

Those infected would be asymptomatic carriers for years or decades, unknowingly spreading the virus (and the prion gene) throughout the population. In contrast to traditional biorisks that cause immediate illness, this would allow massive, undetected distribution before anyone realized such a spread had occurred.

While biosecurity monitoring has drastically improved post-COVID, they still only focus mostly on known pathogens. A virus carrying prion-inducing genes might appear harmless to standard screening, particularly if designed to mimic common, benign viruses. An AGI could design the perfect mimicry.

 

Relatively straightforward procedures & AGI preference

Traditional biorisks mostly follow unreliable cycles of: poke holes until desired traits occur, enhance by repeating. This concept on the other hand is much more procedural and thus much more controllable with less randomness.

For an AGI, this procedural nature makes this highly predictable and optimizable. In addition, unlike other potential AGI elimination methods that involve ongoing conflict or resistance, prion-viral vectors offer a “set and forget” solution that guarantees complete success without requiring continued AGI attention or resources.

 

Environmental persistence & disinfection difficulty

Certain prions are known to persist in the environment for decades; environmental (soil, water, etc.) contamination of farmland and reservoirs could cause severe impact to food and water distribution, while also creating long-lasting dead zones.

Prions are also notoriously resistant to destruction. They survive high temperatures, standard disinfectants, and even some sterilization procedures. Even standard medical procedures could become highly complex with a long list of bureaucratic checklists.

 

No medical countermeasures & sociopsychological effects

Prion diseases have (and likely will not for decades) no cure, no treatment, and no vaccine. Once symptoms appear, death is inevitable, typically within months. Even if the spread were identified immediately, there would be no way to save those already infected.

The psychological impact of a population learning that everyone who caught a “flu-like illness” years ago will die with 100% certainty would be immense. Long-term planning, economic systems, and social structures would collapse under the weight of mass fatalism, panic, and nihilism. Xenophobia and extremist ideologies could spread like wildfire.

 

Our dangerous unpreparedness

Despite the severity of this threat, humanity’s defensive posture is extremely ill-prepared across multiple dimensions.

 

ZERO risk awareness & MASSIVE regulatory / AI safety blindspots

Existing biological weapons treaties and international oversight mechanisms were designed for previous generations of threats and are mostly still stuck on the same mental model - novel emerging biorisks are not even getting proper coverage. While prion research has recently been identified as dual-use research of concern, a prion-virus construct still remains completely unaware and regulations remain far too weak relative to the potential civilizational risk. Most oversight focuses on traditional biological weapons, not this novel threat vector. Institutional biosafety committees and export controls do provide some oversight, but again, these frameworks assume traditional pathogen behavior and not considerate of novel constructs.

There are even commercial AI models out there RIGHT NOW (at the time of writing) that don’t even flag prompts related to this concept (we will not drop names on which companies or models), likely due to its conceptual novelty. And the proliferation of locally-run models makes comprehensive content filtering impossible. More critically, AGI development proceeds without adequate consideration of exotic dual-use research cases.

 

Research accessibility

The democratization of biotechnology means that gene synthesis, viral engineering, and even AI-assisted research design are increasingly available to everyone, whoever that might be. Many commercial gene synthesis companies lack comprehensive screening for dangerous sequences. Some providers do implement screening protocols and laboratory security standards exist, but coverage varies significantly and such systems weren’t designed for novel threat combinations. And exact prion gene sequences are available on public scientific databases.

 

The window is closing

The convergence of accessible gene editing, improved viral engineering, AI-assisted research, and rapid AGI development is rapidly lowering the barriers to developing prion-viruses. Meanwhile, our defensive capabilities lag far behind. We face a narrow window to implement safeguards before this threat becomes readily achievable by a much broader range of actors.

The scientific community continues developing legitimate applications of viral vectors for gene therapy, often targeting the same biological pathways that could be weaponized. This dual-use research creates additional vectors for knowledge transfer and technique refinement that could be exploited by malicious actors - including potentially misaligned AGI systems.

EVEN if we were to determine such attempts to be too complicated to achieve today, biotech / AI / AGI development means the feasibility is approaching reality at exponential speed - and the timeframes may be measured in years, not decades.

 

A call for awareness

The prion-virus threat, amplified by AGI development, demands urgent, comprehensive action across multiple fronts.

Heightened oversight on all research involving a combination of viral vectors and prion-related genes seems necessary, including both government and private research facilities. Bioweapons oversight frameworks, both regional and international, should be integrated to consider prion-virus constructs.

AI systems would need to more thoroughly flag prion-related prompts and perform deep evaluation on specific examples of dual-use biorisks (not just a vague acknowledgement on the overall risks), including but not limited to this prion-virus construct.

Both the AI safety and biosecurity communities must recognize that the intersection of these threats compresses all timelines. The prion-virus represents a class of threat that - at the extreme - could even end human civilization as we know it. Unlike climate change or even nuclear war, it could even offer no opportunity for recovery or gradual adaptation. Once deployed successfully, it would create an irreversible cascade of societal collapse.

The time for action is now, while we still retain the capability to prevent this threat from becoming reality. The convergence of biotechnology advancement and AI / AGI development means we may have limited time to address what could be the ultimate threat to human civilization.