As we noted, the integration of artificial intelligence (AI) into biosurveillance and biodefense architectures to strengthen verification and enforcement mechanisms associated with the Biological Weapons Convention (BWC) can also enable state and non-state actors to obscure, circumvent, or strategically exploit the very compliance frameworks that AI is intended to enhance.  The capacity for AI to model what biosurveillance systems can detect, and under what conditions, creates a range of opportunities for strategic deception. Thus, adversarial actors may leverage AI to design biological agents, their production methods, and distribution pathways that intentionally evade detection thresholds or mimic natural biological variability. The paradox is clear: AI that identifies anomalies can also be trained to eliminate or obscure them.

Public health systems become both targets and unwitting vectors within this paradoxical dynamic. AI-augmented biosurveillance networks that are critical for early outbreak detection produce data streams, that if compromised, could be exploited to mask biological incidents or manipulate epidemiological interpretations. Adversaries could engage AI to generate false public-health data, obscuring an outbreak’s origin or progression, thereby impeding response coordination. Intentionally inaccurate or contrived data could also be propagated to raise false alarms of an outbreak for purposes of creating inter-institutional discord, societal mistrust and instability, time and resource expenditures, and/or focus from other activities.

For the military, this ambiguity imparts significant operational risk. Vagaries in attribution can degrade decision authority, complicate proportional response options, and establish advantage for adversaries to operate below thresholds of overt definition of biowarfare. In expeditionary or multinational operations, such contested information could weaponize confusion and overwhelm intelligence, surveillance, and both medical and command-and-control functions. The resultant erosion of trust in surveillance systems, health reporting, and response capability poses strategic liability.

Politically, this creates uncertainty in BWC governance and international security. States may resist adopting AI-based compliance and verification mechanisms in light of concerns about data sovereignty, espionage, and/or algorithmic integrity or opacity. Conversely, states already operating at the technological margins may exploit extant asymmetries in AI capability to advance their bioagent research and development programs under a guise of compliance. These possibilities further complicate verification diplomacy. Fabricated or distorted biosurveillance evidence could be leveraged to falsely accuse competitors of BWC violations, and such information operations could precipitate political crises, justify military escalation, and destabilize global power balances.

Mitigating the Risks

To preserve the integrity of the BWC while harnessing AI’s benefits, we opine that military and policy leadership must pursue a multilayered mitigation strategy to include:

• Strengthening AI transparency and explainability in order to provide interpretable outputs and standardized metrics to ensure that assessments are trusted by international stakeholders and resilient against adversarial manipulation.
• Developing frameworks for evaluating military AI biosurveillance systems for vulnerabilities against tactics an adversary might employ to evade or corrupt them.
• Integrating public-health and defense intelligence networks and functions to establish broader and more granular (dual-domain) data fusion to: enhance situational awareness, and [b] reduce the likelihood that manipulated indicators of BWC deviation go undetected.
• Advancing international norms for AI in BWC surveillance and oversight to: enable standardized data sharing, auditing, and AI model validation, and [b] reduce geopolitical mistrust toward reinforcing collective compliance.

Toward these goals, we offer the following recommendations:

1. Development of AI countermeasures as a core biodefense domain and mission. Establishing counter-AI analytic systems that can identify adversarially biological designs will be essential. This includes developing models that detect ambiguous genomic features, synthetic signatures, and/or anomalous biological (structural and/or functional) traits.

2. Fortifying multidimensional intelligence capabilities, activities and integration. We opine that biointelligence cannot effectively operate in isolation, but must integrate signals intelligence, cyber forensics, supply-chain analytics, and other forms of biosurveillance (e.g.- waste water analyses, soil and air sampling; etc.) to detect sophisticated, AI-enabled biothreats).

3. Investing in red-, blue- and white-team simulation and wargaming. AI-based multi- teaming will be critical for developing readiness for adversaries’ leveraging novel (cyber-based) methods and tactics for avoiding detection of bioweapon research and development, or interfering with biosurveillance data streams.

4. Promoting international regulations and governances. We have long argued that the regnant scope and tenor of the BWC is insufficient to effectively guide and govern capabilities in bioagent research and production. Enhancing the BWC for the AI era will require more specific and stringent regulations for responsible AI use in the life sciences; [b] establishment of more secure methods and oversight of DNA synthesis and screening powered by advanced AI classifiers; [c] creation of standards for restrictions of high-risk models and/or dual-use computational platforms; and [d] an enforceable system of formalized responses to violations of such restrictions.

Conclusion

We believe that the imperative is clear; if AI systems are to be developed, integrated, and employed to surveille and support the BWC, it is essential to acknowledge both the promise and perils that such systems can afford. Their capacity to enhance detection, verification, and response must be matched by equal vigilance toward the ways they can be manipulated to obscure activity, distort data, and undermine trust. Moving forward, safeguarding the integrity of the BWC will require deliberate, collaborative, and technically rigorous efforts to ensure that AI fortifies the global framework for biosecurity and biological arms control.

Disclaimer

The views and opinions expressed in this essay are those of the authors and do not necessarily reflect those of the United States government, Department of War or the National Defense University.

Dr. Diane DiEuliis is an Expert Consultant of the Institute for National Strategic Studies at the National Defense University.

Elise Annett is Institutional Research Associate at the National Defense University.

Dr. James Giordano is Director of the Center for Disruptive Technology and Future Warfare of the Institute for National Strategic Studies at the National Defense University.