Anthropic's safety hiring focuses on mitigating risks related to nuclear and biological threats.

      An examination of Anthropic's safety hiring reveals its main concerns: they are recruiting analysts to prevent their models from assisting in the creation of nuclear, chemical, and biological weapons. While most job advertisements promote a mission, Anthropic's resemble a risk analysis.

      The company has recently advertised multiple positions for enforcement analysts tasked with ensuring their AI does not aid in weapon development, scams, or cybercrimes, as first reported by Axios. One position seeks an "Enforcement Analyst focused on Radiological & Nuclear Harms," while others address chemicals, explosives, and financial fraud.

      Salaries are in the mid- to upper-$200,000 range, and the roles do not involve coding. Anthropic requires real-world knowledge in areas such as biology and explosives, as well as individuals capable of thinking like attackers aiming to breach its security measures.

      The straightforward job titles are intentional. A spokesperson stated, “Ensuring our models don’t provide potentially harmful information is central to responsible development.” The company frequently hires experts from sensitive domains to evaluate its models before public release.

      Clearly identifying the specific dangers helps attract suitable candidates. Anthropic claims to have hundreds of employees focused on safety, identifying vulnerabilities and addressing them.

      Critics refer to Anthropic as the industry's foremost alarmist, and the pattern of its safety recruitment serves as a response to that criticism. The company is investing substantial resources into the risks it continuously highlights.

      CEO Dario Amodei has dedicated months to outlining potential dangers. In a January essay, he expressed that biological attacks are his greatest concern. “I do not think biological attacks will necessarily be carried out the instant it becomes widely possible,” he wrote. “But accumulated across millions of people and a few years' time, I believe there is a significant risk of a major attack, with casualties potentially reaching millions or more.”

      He has also voiced concerns regarding AI's potential to assist cybercriminals and bolster authoritarian regimes. Earlier this year, Anthropic parted ways with the US Defense Department over the use of its technology for mass surveillance and autonomous weaponry.

      Similar efforts are seen at OpenAI, which is also hiring a researcher focused on biological and chemical risks, offering a base salary up to $445,000. As models become more advanced, all serious laboratories are competing to assemble red teams.

      This competition occurs in an environment devoid of comprehensive AI safety legislation in the US. Congress has been attempting to address this issue for years without success. Some individuals advocate for a regulatory body, with Google’s Demis Hassabis proposing a Wall-Street-style watchdog for advanced models. Less than one percent of AI PhD graduates pursue government roles, leaving expertise largely within private companies.

      The outcome is a peculiar form of self-regulation, where those developing the most dangerous capabilities are also the ones determining how to manage them. Amodei has highlighted this conflict, stating that AI companies pose a risk tier just below hostile nations. His careers page embodies both an argument and a warning: the individuals best equipped to prevent a catastrophe are employed by the company that could potentially instigate it.