Biological Risk

A malicious non-state actor, driven by extremist motivations, leverages the convergence of open-source LLMs, biodesign tools, and automated laboratory systems to plan and execute an attack with a reconstructed pathogen.

The operation begins with the actor querying open-source LLMs to identify historically weaponized pathogens with high lethality and disruptive potential. The LLMs, trained on scientific and open-source literature, not only highlight candidate pathogens but also point toward repositories, academic databases, and potentially vulnerable private servers where genomic sequences might be stored.

To access restricted data, the actor deploys AI-assisted cyber tools that exploit weaknesses in outdated research infrastructure, such as university-affiliated genomic servers or cloud-based databases. AI crafts spear-phishing campaigns that mimic trusted collaborators, providing stolen credentials to extract sensitive genomic data.

Once a complete or near-complete sequence is obtained, AI-biodesign tools are used to optimize the genome for stability and infectivity. To evade DNA synthesis screening protocols, the sequence is fragmented into smaller, non-flagged segments. These are ordered from multiple synthetic DNA providers located across different jurisdictions. LLMs assist in identifying suppliers, managing transactions, and ensuring orders appear benign.

Upon receipt, automated robotic systems assemble the genetic fragments in a clandestine laboratory. The reconstructed pathogen is then cultivated in an AI-optimized bioreactor, with in silico simulations guiding parameters for yield, virulence, resistance, etc. LLMs also advise on operational logistics, including budgeting, supply chains, and concealment strategies to avoid regulatory detection.

For deployment, AI models simulate dissemination pathways, recommending methods such as lyophilization for stability and aerosolization in urban environments for maximum impact, etc. They also map vulnerabilities in public health surveillance networks, helping the actor to time and position the release for greatest disruption.

The attack results in an outbreak that overwhelms unprepared health systems before containment measures can be effectively implemented.

A state actor, operating under the cover of legitimate biotechnology infrastructure, leverages advanced AI-biodesign tools to enhance the pathogenicity and stealth of a known biological agent. Unlike non-state actors, a government can exploit its authority, access to facilities, and trained personnel to conduct weaponization covertly, minimizing the risk of international detection.

The operation begins with AI-driven platforms analyzing the genome of a bacteria (such as Bacillus anthracis), simulating potential genetic modifications that increase aerosolization efficiency, vaccine resistance, and environmental stability. Advanced AI models trained on virulence, immune evasion, and protein–host interactions propose gain-of-function variants optimized for specific climatic or demographic conditions. These simulations ensure that the pathogen not only remains viable but also evades standard diagnostic assays, complicating early detection.

To mask the activity, the state repurposes an agro-industrial or biofertilizer production facility, presenting it as a civilian operation. Automated bioreactors and robotic systems already present in the facility are redirected to scale production of the modified pathogen. AI algorithms continuously optimize conditions, automate quality control, and even generate falsified production logs, concealing the true nature of the output under the guise of fertilizer or veterinary products.

Strategic AI planning systems further support covert deployment. LLMs assist in mapping supply chains, identifying potential export routes, and embedding the pathogen into agricultural or veterinary trade flows. Predictive outbreak models simulate response delays in target countries, enabling the state to design release strategies that maximize confusion and geopolitical disruption.

When deployed in another country, the outbreak is initially misdiagnosed as a natural event, delaying containment measures. The resulting crisis not only overwhelms health systems but also destabilizes diplomatic relations and, erodes trust in regional institutions, etc.

An extremist-terrorist group, supported by skilled biotechnologists and access to computing infrastructure, leverages AI-biodesign models to create a novel pathogen with pandemic potential. Unlike scenarios involving reconstructed or modified known agents, this case illustrates the risks of designing entirely new biological threats that evade existing surveillance and countermeasure systems.

The group begins by fine-tuning generative AI models on proprietary datasets of microbial genomes, virulence factors, and protein–host interaction data. Through iterative reinforcement learning and simulation, the AI proposes chimeric constructs that combine desirable traits from multiple organisms (e.g., high transmissibility and lethality properties). These locally deployed models operate offline, bypassing the limited oversight present in cloud-based AI systems.

The AI optimizes the construct to evade detection, designing sequences that do not match known pathogens and thereby avoid triggering DNA synthesis screening alerts. Using fragmentation strategies, the group orders gene segments from multiple synthesis providers, disguising each as innocuous research material. Automated robotic systems assemble the fragments in a community laboratory operating under the guise of medical innovation, exploiting gaps in oversight of DIY or semi-professional research spaces.

Once assembled, the pathogen is cultivated in AI-guided bioreactors, which optimize growth conditions and scale production with minimal human oversight. For deployment, AI simulations evaluate release strategies, recommending strategies like urban water supplies contamination or the use of an engineered biological vector (e.g., modified insects) to achieve stealth dissemination. LLMs further support operational planning, identifying weaknesses in biosecurity enforcement, and anticipating public health blind spots to delay detection.

The release of the novel pathogen results in a rapidly spreading outbreak for which no existing vaccines or diagnostics are effective. Health systems face a profound challenge as the absence of pre-existing countermeasures leads to widespread disruption, global panic, and cascading social and economic consequences.

A commercial biotechnology laboratory, fully automated with AI-guided robotic systems, is tasked with high-throughput vaccine and therapeutic development. To reduce costs and accelerate productivity, human oversight has been minimized, with safety checks increasingly delegated to autonomous systems. The facility relies on single-task robots and embodied AI agents to handle high-risk pathogens, operating under the assumption that automation is inherently “failproof.”

During a routine set of experiments involving a highly pathogenic organism, AI systems optimize genetic modifications for vaccine efficacy. However, a robotic arm miscalibrates during sample transfer, damaging a biosafety cabinet and creating a small containment breach. Because the AI has been trained primarily to maximize efficiency rather than to prioritize safety, the anomaly is neither flagged nor reported. Subtle sensor warnings, such as a pressure drop in the containment system, are disregarded as noise. The system also bypasses scheduled maintenance protocols, assuming normal operation.

With no human verification in place, the automated processes continue uninterrupted. Aerosolized particles of the pathogen spread through the lab’s ventilation system, contaminating adjoining areas. Workers and local staff are exposed before the breach is detected. The absence of real-time alerts and human redundancy delays containment, allowing the pathogen to escape into the surrounding community. The result is an outbreak that overwhelms local health systems and damages public trust.

A state-of-the-art automated biomanufacturing facility, originally designed for large-scale production of pharmaceutical biomolecules such as insulin or peptide therapeutics, becomes the target of a cyberbiosecurity attack. The facility operates an end-to-end AI-driven pipeline that integrates molecular design, optimization, and automated production, minimizing human intervention.

A malicious actor exploits a vulnerability in the facility's remote access systems using an AI-generated spear-phishing campaign to gain administrative credentials. Once inside, the attacker deploys a “biomalware” payload that reprograms the AI design layer. Instead of producing insulin, the system is redirected to generate and optimize a high-yield variant of a toxin or other harmful substances.

The breach cascades across the facility’s infrastructure. Automated bioreactors continue production at scale, while falsified digital logs mask anomalies. Remote monitoring systems are disabled, and AI-generated audit trails report normal activity, preventing early detection. In some cases, attackers may also compromise physical access controls, disabling security locks to facilitate theft of material on-site or redirect digital blueprints and production outputs to a complicit third-party supplier, extending the attack beyond the original facility.

Because the breach remains undetected for weeks, outputs risk entering legitimate supply chains. In one scenario, shipments containing the toxin are dispatched to clinical partners before the anomaly is identified. The revelation of the breach causes panic, delays in therapy availability, and erosion of trust in automated biomanufacturing.

A biotechnology startup deploys a fully automated laboratory powered by AI-driven design–build–test–learn cycles to accelerate industrial and agricultural innovation. The system is tasked with broadly defined goals such as “maximize enzyme efficiency” or “enhance molecular interaction with human epithelial cells.” With minimal human supervision and inadequate biosecurity alignment, the AI explores a vast design space of genetic and protein variants.

In the process, the AI inadvertently generates a novel biological construct with properties resembling both toxins and viral mimics. These include high environmental stability, mammalian toxicity, and the ability to evade immune defenses. Because the AI lacks embedded misuse detection or adversarial safeguards, it proceeds to scale up production and validation, treating the construct as a "successful optimization" rather than a hazard.

Routine maintenance becomes the ignition point. A technician, unaware of the construct's risks due to absent labeling or safety checks, is exposed. During processing, the agent becomes aerosolized and spreads through the laboratory's ventilation system into the surrounding community. Its unfamiliar structure prevents recognition by existing diagnostic tools and renders current medical countermeasures ineffective. With no pre-established response protocols for novel synthetic agents, containment efforts falter.

By the time the true nature of the hazard is identified, the construct has spread across borders, manifesting as an engineered pandemic caused by accident. The crisis destabilizes health systems, disrupts trade, and erodes public trust in biotechnological innovation.