"It has long been a concern that AI models might one day reach the level of capability where they can help to make existing pathogens more dangerous—or create entirely new ones," Claude maker Anthropic recently said. This was part of the AI firm's recent announcement of how it blocked efforts by certain "threat actors" who wanted to use its artificial intelligence models for a slew of questionable activities, including cyberattacks, surveillance, and, more dangerously, research on biological weapons.
According to Anthropic, AI models getting more powerful means that cyberattacks no longer need sophisticated hacking skills. Even a single person with enough computing power can create cybersecurity threats that were not possible a year ago, Anthropic said.
The company said it has added stronger safeguards in its latest models to restrict biological research that could also be used to make weapons.
Therefore, "out of an abundance of caution", the AI firm said that its latest models, such as Claude Fable 5, have "stronger safeguards that restrict access to a wide range of dual-use biological research queries".
However, it did state: "To our knowledge, no private company, AI or otherwise, has yet shared evidence of the potential misuse of their platforms for biological weapons development publicly."
One such attempt involved "research that genetically alters an organism to create a new or enhanced biological property" on the chikungunya virus, it said. It is not merely the research that triggered the red flag.
It was who triggered it. Anthropic explained: "The countries conducting this research are in regions where Anthropic does not supply service, so the platform tunnelled traffic through US infrastructure to evade our regional blocks, and used a zero data retention (ZDR) service to hide content. The developers of this platform used Claude through grey market resellers and synthetic accounts outside the ZDR channel to support their work, and explicitly referred to academic researchers as customers who were sensitive to the blocking actions of our safety classifiers."
Another case involved research on highly pathogenic avian influenza, or bird flu. "The researcher in question accessed Claude from an unsupported region via US virtual private server infrastructure, using a privacy-email provider with an auto-generated username," it explained.
And another case involved someone using an account "created from a randomly generated email address shortly before use and operated through anonymising US infrastructure, with operator logins traced to proxies shared with a banned account farm" who looked for access to "high-containment facilities and planned work with live orthopoxviruses". Variola, the agent of smallpox and Mpox, is an orthopoxvirus.
This was Anthropic's third such report since March that detailed the misuse of AI. "The cases we share here aren't typical misuse, but rather examples of the most notable and novel threat activity we've identified to date," it stressed.
The timing of the report, however, is highly suspect, according to AI ethics watchers. Two days before its publication, one of Anthropic's researchers announced that he resigned after alleging that the AI firm and its peers were not acting responsibly with AI development. He also detailed how the technology had the potential to eventually evade human control.
Adding to that is how generative AI models are somehow individualised in commonspeak. "It is always Claude did this, ChatGPT did that. We often forget that there are big corporations run by people who benefit from their use. They need to be responsible for it," stated 37-year-old Kishan R, an AI researcher based in Noida.
"It is instances like this where we need to explore the ethics of it. How are companies like Anthropic and xAI determining what constitutes safe use of their AI models? This is a democratic question and not a corporate one. Such rules should be determined by a regulatory body with oversight and powers," the researcher added.
