AI Safety Groups Face a Talent Shortage

Rishub Jain left Google DeepMind’s AI safety team to cofound Sampura Research, a nonprofit where he says he can pursue research on an AI alignment issue that he felt had been deprioritized. He said major AI companies invest in safety but could invest more; Google did not comment on whether it continues to focus on that specific research area. DeepMind’s principles describe a holistic approach to responsibility, safety and security.

AI safety job openings on LinkedIn rose 91% over the past year. Researchers and advocates say the field faces a shortage of people, including specialists in technical research, policy and governance. Laura González Salmerón of 80,000 Hours said, “Talent is the main bottleneck.” The article says the industry expects a surge of philanthropic money tied to anticipated IPOs at Anthropic and potentially OpenAI.

The competition for workers spans large companies, nonprofits and startups. Sampura lists $273,000 to $395,000 for an experienced team leader, a salary the article says would place the worker in the top 3% of US earners. Meta reportedly offered $100 million in signing bonuses last year to recruit leading experts; OpenAI employees receive an average of $1.5 million in stock-based compensation, while Anthropic technical staff can earn more than $1 million annually. Researchers disagree over whether they can reduce risks more effectively inside companies or from outside them.

Jain said he worries that AI development may outpace efforts to align systems with human values. David Robinson, who led transparency work on OpenAI’s safety team, wrote that companies need more layers of checks, comparing them with safeguards in airports and nuclear energy. He said staff were too busy sprinting to consider or make fundamental changes, and plans to work on the issues outside the company. OpenAI later fired three researchers, saying they improperly shared sensitive internal data with third-party safety researchers. The researchers argued in a letter to the board that their engagement was within their roles and was chilling remaining staff; OpenAI said the firings were not about raising safety concerns or speaking out.

Training and recruitment groups are also drawing interest from people outside AI research. Kairos co-director Neav Topaz said he speaks with about 10 to 20 prospective career changers each month. The nonprofit raised $50 million from Coefficient Giving and launched its Generator Residency this summer: 27 people trained for three months to transition into AI safety, combining expertise such as consulting with safety principles. A dozen residents have already secured roles, and many others have offers, Topaz said.

Funding and public attention have expanded in 2026. 80,000 Hours has spoken with about 8,000 people since it was founded in 2011, with most calls focused in some way on AI, González Salmerón said. METR said it raised $71 million in six months this year. AI safety groups and individual donors spent nearly $50 million this year on election candidates focused on AI regulation, Axios reported, while Jaan Tallinn has donated $170 million to AI safety nonprofits.

Coefficient Giving’s Project Tailwind, announced in September, says it is on track to give more than $1 billion this year to help companies prepare for an expected increase in philanthropic capital. Sampura received an $11 million grant from the group. Coefficient Giving’s Rossa O’Keeffe-O’Donovan said gaps remain in the safety ecosystem and described a need for people who combine technical expertise, risk analysis and startup skills. Outside organizations may face limited access to computing power and leading AI models, less influence and lower pay, while potential roles also include consulting, policy, communications and business leadership.

Rishub Jain left Google DeepMind’s AI safety team to cofound Sampura Research, a nonprofit where he says he can pursue research on an AI alignment issue that he felt had been deprioritized. He said major AI companies invest in safety but could invest more; Google did not comment on whether it continues to focus on that specific research area. DeepMind’s principles describe a holistic approach to responsibility, safety and security.

AI safety job openings on LinkedIn rose 91% over the past year. Researchers and advocates say the field faces a shortage of people, including specialists in technical research, policy and governance. Laura González Salmerón of 80,000 Hours said, “Talent is the main bottleneck.” The article says the industry expects a surge of philanthropic money tied to anticipated IPOs at Anthropic and potentially OpenAI.

The competition for workers spans large companies, nonprofits and startups. Sampura lists $273,000 to $395,000 for an experienced team leader, a salary the article says would place the worker in the top 3% of US earners. Meta reportedly offered $100 million in signing bonuses last year to recruit leading experts; OpenAI employees receive an average of $1.5 million in stock-based compensation, while Anthropic technical staff can earn more than $1 million annually. Researchers disagree over whether they can reduce risks more effectively inside companies or from outside them.

Jain said he worries that AI development may outpace efforts to align systems with human values. David Robinson, who led transparency work on OpenAI’s safety team, wrote that companies need more layers of checks, comparing them with safeguards in airports and nuclear energy. He said staff were too busy sprinting to consider or make fundamental changes, and plans to work on the issues outside the company. OpenAI later fired three researchers, saying they improperly shared sensitive internal data with third-party safety researchers. The researchers argued in a letter to the board that their engagement was within their roles and was chilling remaining staff; OpenAI said the firings were not about raising safety concerns or speaking out.

Training and recruitment groups are also drawing interest from people outside AI research. Kairos co-director Neav Topaz said he speaks with about 10 to 20 prospective career changers each month. The nonprofit raised $50 million from Coefficient Giving and launched its Generator Residency this summer: 27 people trained for three months to transition into AI safety, combining expertise such as consulting with safety principles. A dozen residents have already secured roles, and many others have offers, Topaz said.

Funding and public attention have expanded in 2026. 80,000 Hours has spoken with about 8,000 people since it was founded in 2011, with most calls focused in some way on AI, González Salmerón said. METR said it raised $71 million in six months this year. AI safety groups and individual donors spent nearly $50 million this year on election candidates focused on AI regulation, Axios reported, while Jaan Tallinn has donated $170 million to AI safety nonprofits.

Coefficient Giving’s Project Tailwind, announced in September, says it is on track to give more than $1 billion this year to help companies prepare for an expected increase in philanthropic capital. Sampura received an $11 million grant from the group. Coefficient Giving’s Rossa O’Keeffe-O’Donovan said gaps remain in the safety ecosystem and described a need for people who combine technical expertise, risk analysis and startup skills. Outside organizations may face limited access to computing power and leading AI models, less influence and lower pay, while potential roles also include consulting, policy, communications and business leadership.