Anthropic, the company behind the Claude AI model, is significantly expanding its security teams by opening 32 specialized positions aimed at preventing the malicious use of artificial intelligence. These roles cover a broad spectrum of threats, ranging from explosives manufacturing and nuclear weapon development to cyberattacks and biological terrorism.
From Nuclear Threats to Red Teaming
The company is seeking analysts and researchers with expertise in highly sensitive fields, specifically CBRN-E threats (chemical, biological, radiological, nuclear, and explosives). These specialists will employ "red teaming" methodologies, attempting to think like malicious actors to identify system vulnerabilities before public release.
According to Anthropic, their work involves:
- Bypassing safety filters to identify loopholes.
- Ensuring models do not provide dangerous technical information.
- Testing if the AI can synthesize innocent data into hazardous instructions.
High Salaries and the Privatization of Safety
The specialized knowledge required is rewarded with annual salaries ranging from $200,000 to $300,000. This trend highlights a burgeoning job market at the intersection of technology and national security. However, the concentration of such expertise within private corporations raises questions about who ultimately decides which risks are acceptable.
Conflicts and Industry Competition
Anthropic CEO Dario Amodei has repeatedly warned about the risk of AI-facilitated biological attacks. Meanwhile, the company has reportedly clashed with the U.S. Department of Defense over the potential use of its technology in mass surveillance systems. Simultaneously, OpenAI is following a similar path, offering salaries as high as $445,000 for comparable biosafety roles.
