5 Roles That AI Has Created or Reshaped in 2026

'AI is creating new jobs' is easy to say and hard to document. Most evidence about work is slow, aggregated and cautious, while job-title lists on social media are fast and vivid. This article stays with the slow, checkable sources: US government projections, a hiring-data analysis, institutional pages and an international labour-agency brief. We read each on 18 October 2026 and say what each does and does not show.

How did we choose these five roles?

We looked for roles where a primary or institutional source we could read says something specific. We excluded anything that rested only on a search summary or a blog quoting an unseen report. That is why the list is not the most exciting one possible. For each role we label the evidence as an official projection, a sample of postings, or an institutional description.

1. Data scientist: an old job pulled forward by AI?

The US Bureau of Labor Statistics (BLS) Occupational Outlook Handbook states that employment of data scientists is projected to grow 35 percent from 2025 to 2035, which it calls much faster than average, with about 24,800 openings per year on average. Its page says firms are expected to continue to integrate AI-based systems into their workflows, alongside growing demand for data-driven decisions. The page was last modified on 27 August 2026.

New or reshaped: reshaped. Data science is an established occupation. What it does not show: BLS does not say how much of the growth comes from AI as opposed to other uses of data, and many of the openings are replacements for people who leave the field. Projections can also be wrong.

2. Computer and information research scientist: who builds the models?

BLS projects 22 percent growth in employment for computer and information research scientists from 2025 to 2035, about 2,900 openings per year. It states that computer scientists' expertise will be needed in the creation of new technologies related to AI, and also cites data-mining demand and cybersecurity as drivers.

New or reshaped: reshaped. This is a long-standing research occupation. What it does not show: it is a small occupation, so a high growth rate means relatively few absolute jobs, and the page does not separate AI research from other computer science research.

3. AI governance specialist: what does the hiring data say?

Axial Search, a recruiting firm, analysed 1,997 US AI governance postings from online job boards, covering roles posted since January 2026, in an article updated on 4 August 2026. It reports that 73% of postings sit at senior level or above, with manager roles at 28% and individual-contributor roles at 46%. Professional services leads with 35% of postings, with technology and financial services at 13% each. California accounts for 14% of postings and New York 13%.

New or reshaped: newer in name. Compliance, risk and policy work existed before AI, and this analysis counts postings that match AI governance, not first-ever job titles. Limits the author states: skills and leadership profiles are inferred from posting text and AI analysis, and a missing item means it was not stated, not that it is unwanted. It covers US postings only. We quote no salary figures because they derive from posted pay bands reported as percentiles by a firm with an interest in the market.

For why governance matters in practice, see our account of AI safety incidents and warning signs of 2026 and our guide to procurement evidence.

4. AI model tester or evaluator: is this a real job yet?

The UK government's AI Security Institute describes itself as a research organisation within the Department for Science, Innovation and Technology and lists 'testing leading AI systems' among its work, saying it tests systems before and after they are launched. Its page mentions more than 100 technical staff. It does not use the term red-teaming and does not tie specific staff to specific tests.

New or reshaped: newer in practice, though the title varies by employer. What it does not show: the page gives no job counts, no pay and no hiring trend, and it shows no date. We found no readable 2026 posting data for evaluator or red-team roles, so we cannot say how large this job market is. Our guides on reading benchmark claims and how to read an evaluation explain why independent testing matters to buyers.

5. AI data worker: the labour behind the models?

The Oxford Internet Institute's Fairwork project rates working conditions in the platform economy against five principles: fair pay, fair conditions, fair contracts, fair management and fair representation. Its site describes 'cloudwork' as online work that can be done from anywhere, including 'filling in surveys and training AI systems', and it lists a 2026 evaluation of an AI supply-chain company among its publications. Companies are assessed with the same standards and can submit evidence.

New or reshaped: a growing form of online labour that predates the current wave of AI but is tied to it. What it does not show: we could not open Fairwork's full reports (the Appen report PDF exceeded our reader's size limit), so we state none of their scores or findings. Journalistic reports of low pay and difficult conditions for annotators exist, but we did not read them in full and do not repeat their figures here. If you work in this field or hire people in it, read the original reports.

What is AI reducing or changing?

BLS projects that employment of customer service representatives will decline 5 percent from 2025 to 2035. Its page says demand is expected to fall as tasks continue to be automated, with self-service tools, social media and apps letting customers handle simple requests. The page does not name AI as the cause, so we should not either.

The International Labour Organization's May 2025 update on generative AI and jobs says one in four workers worldwide are in an occupation with some degree of generative AI exposure, and that most jobs will be transformed rather than made redundant because human input remains necessary. It assessed occupations at the six-digit level using nearly 30,000 tasks. It also notes that growing abilities in voice, image and video have raised automation scores for some media and web occupations.

How can you tell a new job from a relabelled one?

What should a worker or employer do with this?

• Workers: look at the tasks in your role that AI tools already change, and learn to check AI output. Our guide to useful AI prompts is a starting point for the checking habit.
• Employers: be wary of promises to replace whole roles. Evidence on what works is still thin, as our vendor claims vs verified findings guide explains.
• Students: read the primary pages we cite. Government outlook pages are updated and are free to read.

What are the limitations of this article?

• It covers five roles, chosen because we could read their sources, not because they are the largest or fastest growing. It is not a ranking.
• Most figures are US-centric. The ILO brief is global, and the UK institute page is a single institution.
• The World Economic Forum's Future of Jobs report was not readable by our tools, so we use none of its figures. Readers should consult it directly.
• Press coverage of specific companies creating new AI job titles is widely repeated, but we could not read the primary sources, so we left it out.
• Pages were read on 18 October 2026 and projections can be revised. See our corrections page and methodology.