biosecurity
In plain wordsProtecting people from the deliberate or careless spread of dangerous germs and toxins; in AI, making sure AI tools do not make it easier to create or release biological weapons.
Policies and practices that protect dangerous biological agents and toxins, and the knowledge and technologies for producing them, against people who would steal them or deliberately cause harm with them; biosafety, by contrast, deals with accidents such as laboratory infections. In AI, biosecurity addresses how language modelslarge language model, n.: A computer program trained on enormous amounts of text to predict the words that come next. At vast scale, that simple skill lets it chat, answer questions, translate, summarize and write code., agentsAI agent, n.: An AI system that does not just answer but acts: given a goal, it decides on steps, uses tools such as a browser or code editor, checks the results and keeps going until the job is done. and biological design tools (for protein structure, sequence design or genome engineering) could lower barriers to producing dangerous pathogens or toxins, and how AI can strengthen defenses. Measures include evaluating models for biological knowledge and upliftuplift study, n.: An experiment that measures how much an AI really helps people do a task — often a dangerous one, like planning a biological attack — compared with people using only ordinary resources such as the internet., refusing or filtering hazardous requests, restricting accessstructured access, n.: Letting people use an AI system through a controlled channel, such as a website or programming interface, instead of handing over the model itself, so that the developer can limit misuse. to the most capable design tools, screening orders to commercial DNA synthesis providers for sequences of concern and verifying customers, and using AI for pathogen surveillance and medical countermeasures. It is the most prominent strand of CBRN riskCBRN risk, n.: The danger that AI could help someone make or use chemical, biological, radiological or nuclear weapons — for example, by explaining hard technical steps an attacker could not work out alone. in AI policy.
The biosecurity review covered both the protein-design model's release terms and its synthesis partners' order screening.
Origin
Long used in agriculture and public health, the word's sense of guarding against deliberate misuse became prominent after the 2001 anthrax letters. AI biosecurity became prominent after Fabio Urbina and colleagues reported in 2022 that a drug-discovery model with its objective inverted proposed about 40,000 candidate toxic molecules in under six hours, and after 2023 studies and policies, including US Executive Order 14110, addressed language models and DNA synthesis screening.
Related
- CBRN riskCBRN risk, n.: The danger that AI could help someone make or use chemical, biological, radiological or nuclear weapons — for example, by explaining hard technical steps an attacker could not work out alone.
- dual-usedual-use, adj.: Describes technology or knowledge that can be used both to help and to harm: the same AI that finds software bugs for defenders can find them for attackers.
- uplift studyuplift study, n.: An experiment that measures how much an AI really helps people do a task — often a dangerous one, like planning a biological attack — compared with people using only ordinary resources such as the internet.
- WMDPWMDP, prop. n.: A multiple-choice test of AI knowledge that could help someone make biological, chemical or cyber weapons, built from questions near the dangerous line but not over it, and used to test ways of erasing such knowledge.
- structured accessstructured access, n.: Letting people use an AI system through a controlled channel, such as a website or programming interface, instead of handing over the model itself, so that the developer can limit misuse.
- misusemisuse, n.: People deliberately using AI to cause harm — for scams, harassment, propaganda, hacking or weapons — as opposed to harm from AI making mistakes or pursuing goals of its own.
Sources
- Urbina, F., Lentzos, F., Invernizzi, C. & Ekins, S. (2022). Dual use of artificial-intelligence-powered drug discovery. Nature Machine Intelligence 4.
Mentioned in
CBRN riskCBRN risk, n.: The danger that AI could help someone make or use chemical, biological, radiological or nuclear weapons — for example, by explaining hard technical steps an attacker could not work out alone., uplift studyuplift study, n.: An experiment that measures how much an AI really helps people do a task — often a dangerous one, like planning a biological attack — compared with people using only ordinary resources such as the internet., differential technological developmentdifferential technological development, n.: A strategy of speeding up technologies that make us safer and slowing down risky ones, so that protections arrive before the dangers they guard against., Virology Capabilities TestVirology Capabilities Test, prop. n.: A hard test of practical laboratory virology know-how, used to check whether AI models know enough about working with viruses that they might help someone misuse biology., vulnerable world hypothesisvulnerable world hypothesis, n.: The idea that some future invention could make mass destruction so easy that civilization would be wrecked by default, unless it had unusually strong ways of preventing misuse., WMDPWMDP, prop. n.: A multiple-choice test of AI knowledge that could help someone make biological, chemical or cyber weapons, built from questions near the dangerous line but not over it, and used to test ways of erasing such knowledge.