AI safety has evolved from a niche academic concern into a global industry. More than 180 organisations now focus on some aspect of AI safety, security, governance, or regulation. As investment accelerates, demand for AI safety talent, governance specialists, and machine learning researchers has grown significantly, creating new challenges for employers seeking to recruit technical expertise.  

The landscape spans seven categories: frontier AI labs, AI security companies, research organisations and non-profits, government bodies, academic institutions, industry consortia, and open-source projects. Together, these groups are shaping how advanced AI systems are developed, evaluated, secured, and governed.

The ecosystem is concentrated in a few hubs, majoring in San Francisco, Washington DC, London, Berkeley, and New York. But AI safety is becoming increasingly global, with significant growing communities in Tel Aviv, Paris, Geneva, Seattle, Singapore, Tokyo, Seoul and Oxford.

Funding has poured into the space. Frontier AI companies have raised billions of dollars, while governments, research funders, and philanthropic organisations are collectively investing hundreds of millions annually in AI safety research, testing, standards development, security capabilities and specialist skills development.

Frontier AI Labs with Safety Teams

The major AI developers have become some of the largest investors in AI safety.

Anthropic, founded in 2021 by former OpenAI researchers, is perhaps the most safety-focused frontier AI lab. Its Constitutional AI approach trains models to be helpful, harmless, and honest using a set of principles rather than RLHF alone. Anthropic's Responsible Scaling Policy establishes capability thresholds that trigger additional safety measures before deploying more powerful models.

OpenAI maintains a Preparedness team that evaluates frontier models across cybersecurity, CBRN (chemical, biological, radiological, nuclear), persuasion, and autonomous behaviours. The Preparedness Framework mandates that models rated "high" risk in any category cannot be deployed.

Google DeepMind works on alignment research, dangerous capability evaluations, and the Frontier Safety Framework, while Meta focuses on open-source model safety, and Mistral AI integrates safety into European-developed models.

A notable new entrant is Safe Superintelligence Inc. (SSI), co-founded by former OpenAI chief scientist Ilya Sutskever. Unlike other AI companies, SSI focuses exclusively on solving the challenge of aligning AI superintelligence with human goals and values - making it the first company dedicated solely to this problem.  

The Rise of AI Security Companies

Alongside the frontier labs, a new generation of specialist AI security companies has emerged.

These companies focus on protecting AI systems from threats such as prompt injection, model theft, adversarial attacks, data leakage, and compromised machine learning supply chains. Examples include HiddenLayer, Lakera, Protect AI, Mindgard, and CalypsoAI.

The sector is developing rapidly because traditional cybersecurity tools were not designed to address AI-specific risks. Organisations deploying LLMs increasingly require specialised capabilities such as automated red-teaming, model monitoring, AI risk assessment, and defensive controls for AI applications.

As demand grows, the market is already consolidating. Established cybersecurity firms including Cisco, Palo Alto Networks, CrowdStrike, and Wiz have expanded into AI security through acquisitions, partnerships, and internal development.

Research Organisations and Non-Profits

Much of the foundational AI safety work still comes from independent research organisations and non-profit institutions.

The Machine Intelligence Research Institute (MIRI), founded in 2000, remains one of the earliest organisations dedicated to long-term AI alignment research. The Centre for AI Safety (CAIS) has become a major coordinator of safety research and policy discussions, while organisations such as METR, Apollo Research, and Redwood Research focus on evaluating advanced AI systems and understanding emerging risks.

Grant-making organizations fund the ecosystem: Open Philanthropy is the largest funder of AI safety research, the Long-Term Future Fund provides targeted grants for safety upskilling, and the Survival and Flourishing Fund supports existential risk reduction efforts.  

Research training programmes such as MATS and AI Safety Camp help develop the next generation of AI safety professionals, creating important talent pipelines for employers. As demand for AI skills continues to grow, these programmes play a critical role in addressing capability shortages across industry, government, and academia.

Governments Take Centre Stage

Perhaps the most significant recent development has been the growing involvement of governments.

The UK's AI Safety Institute, established following the 2023 AI Safety Summit, has become a world-leading centre for evaluating frontier AI models. Similar organisations have since been created in the US, Japan, South Korea, Canada, and France, forming an international network focused on testing and assessing advanced AI systems.

Regulators and standards bodies are also becoming increasingly influential. The EU AI Office is responsible for implementing the EU AI Act, while organisations such as NIST in the US are developing measurement standards, testing methodologies, and risk management frameworks for AI systems.  

Government agencies are not only shaping regulation but are also becoming major recruiters of AI specialists. AI safety institutes, regulators, and national security organisations are expanding their hiring programmes to attract expertise in AI policy, model evaluation, risk management, and technical assurance.

Universities and Academic Research

Universities remain central to the AI safety ecosystem, providing much of the theoretical research that informs industry and government practice.  

In the UK, the University of Oxford hosts multiple AI safety research groups including the Centre for the Governance of AI (GovAI), the Future of Humanity Institute (now closed but historically foundational), and the Leverhulme Centre for the Future of Intelligence (LCFI).  

Cambridge has CSER (Centre for the Study of Existential Risk) and the Alan Turing Institute in London works on trustworthy AI.

Academic researchers continue to make important contributions in areas such as alignment, interpretability, robustness, governance, and AI evaluation methodologies.

Key Focus Areas Across the Ecosystem

Focus areas tracked across the ecosystem reveal where the community concentrates its efforts. Governance (96 organizations) and responsible AI (95) lead – reflecting the industry's emphasis on policy and ethical frameworks. Evaluation (69) and AI policy (65) are next, showing strong investment in measuring safety and shaping regulation.

Technical safety areas include compliance, alignment, fairness, red-teaming), robustness, secure deployment, and LLM security. Monitoring and threat modelling reflect the growing emphasis on operational security.

Specialized areas include interpretability, content safety, guardrails, privacy, supply chain, incident response, and watermarking.

From Niche Discipline to Global Ecosystem

The AI safety ecosystem has matured into a major global sector encompassing research, regulation, security, and commercial innovation. What began as a small community focused on long-term risks has expanded into a broad network of organisations addressing the practical challenges associated with deploying increasingly capable AI systems.

This growth is creating increasing demand for AI talent. From AI safety researchers and AI governance professionals to AI security specialists, organisations across every sector are competing to recruit individuals with the skills needed to develop, oversee, and secure AI systems responsibly.

Find out more about our AI recruitment  and Cybersecurity recruitment solutions.