The new appointment comes in light of concerns that advanced AI systems could be misused to cause catastrophic harm, including by providing guidance related to chemical or radioactive weapons. These concerns have intensified amid the ongoing US-Israel Iran war.

Anthropic joins other AI developers, including OpenAI, in hiring skilled safety and defence experts to mitigate risks associated with their technologies. 

The role commands a significant salary, with Anthropic advertising an annual salary range of $245 – $285k. 

The LinkedIn advert calls for applicants with extensive practical experience, as well as a high level of academic credentials (Ph.D required). 

Candidates are expected to be “comfortable navigating ambiguity and developing solutions for novel safety challenges”, a nod to AIs rapidly-evolving capabilities, with the overall  purpose of the role to ensure “AI systems remain safe and beneficial”. 

Despite Anthropic’s high profile, the niche, high-stakes position appears to be attracting a relatively limited pool of candidates on LinkedIn. 

Experts warn that employing such professionals requires a very careful approach. While assessing risks is essential, organisations must also consider how in doing so, sharing sensitive expertise with AI systems, could unintentionally increase those risks.

Recommended reading

• OpenAI Backtracks on “Opportunistic and Sloppy” US Government Deal
• Anthropic Out, OpenAI In | The US AI Power Struggles
• Anthropic Reaches $380bn as AI Investment Frenzy Rolls on

Dr Stephanie Hare, tech researcher and co-presenter of the BBC’s AI Decoded TV programme commented: “Is it ever safe to use AI systems to handle sensitive chemicals and explosives information, including dirty bombs and other radiological weapons? 

She added: “There is no international treaty or other regulation for this type of work and the use of AI with these types of weapons. All of this is happening out of sight.”

The hiring campaign comes as Anthropic reportedly pursues legal action against the US Department of Defence, following the department’s claims that its AI systems posed a supply chain risk. The company failed to come to an arrangement with the department over supporting certain surveillance uses, sparking the department’s blacklisting. 

As AI capabilities continue to advance, the challenge becomes not just what can the systems do, but who will set the limits of them. Roles like this will become key in defining the AI capabilities of the future.