From Sanskriti Khandelwal | Product & Market Analysis
AI Ethics Teams on the Record: 15 Dated Events at 6 Companies
On this page
Between December 2020 and February 2026, we found 15 dated, publicly reported events affecting AI ethics teams at Microsoft, Google, Meta, Amazon, Twitter (now X) and OpenAI. Thirteen of them were cuts, departures, dissolutions or reassignments. The other 2 were growth disclosures, and both came from Microsoft. Most coverage frames this as companies scrapping their AI ethics teams. The paper trail shows something less tidy and harder to check: teams were mostly moved, not sacked.
Key takeaways
- 13 of 15 public events on corporate AI ethics teams since 2020 were reductions or restructurings. The 2 growth disclosures in the record both came from Microsoft, in May 2023 and May 2024.
- Reassignment, not layoff, is the most common outcome. 5 of the 15 events moved an ethics or safety team into product, trust and safety, or research groups, at Meta, Google and OpenAI.
- Microsoft eliminated one ethics team in the same year it reported about 350 responsible AI staff. Its Ethics and Society team went from roughly 30 people in 2020 to zero in March 2023.
- OpenAI dissolved 3 safety or mission teams in 21 months. Superalignment in May 2024, AGI Readiness in October 2024 and Mission Alignment in February 2026.
Who this is for: founders and directors choosing AI vendors, because a supplier's internal oversight is now part of your own business ethics and procurement risk.
The short answer. Microsoft, Twitter and Amazon cut AI ethics staff outright between 2022 and 2023. Meta, Google and OpenAI mostly dissolved named teams and moved the people into product, trust and safety, or research groups. Only Microsoft has published headcount showing growth. None of the six publishes a complete, dated history of its own ethics teams.
The AI ethics team ledger, 2020 to 2026
Each row below is one event with a date and a source we opened. We counted an event when a named ethics, responsible AI or safety team was cut, dissolved, reassigned, or reported with a new headcount. We also counted the 2 departures of Google's Ethical AI co-leads, because those exits reshaped the team that remained.
We did not count general layoffs that happened to include an ethics researcher. We did not count trust and safety cuts unless reporting tied them to AI ethics work. One adjacent case, Google's Jigsaw unit, is included and flagged, because CNBC reported it alongside the AI ethics cuts.
| Date | Company | Team | What happened | Outcome type |
|---|---|---|---|---|
| Dec 2020 | Ethical AI | Co-lead Timnit Gebru exits. Google called it a resignation. Gebru described it as a firing. | Leadership exit | |
| 19 Feb 2021 | Ethical AI | Co-lead Margaret Mitchell fired. Google cited code of conduct and security policy violations. | Leadership exit | |
| Oct 2022 | Microsoft | Ethics and Society | Reorganisation cuts the team to roughly 7 people, from about 30 at its 2020 peak. | Shrunk |
| Nov 2022 | Twitter (now X) | ML Ethics, Transparency and Accountability | All but 1 member laid off. The director said the team had 17 people. | Eliminated |
| Feb 2023 | Jigsaw (adjacent) | About one third of a unit working on misinformation and toxicity cut. | Shrunk | |
| Mar 2023 | Amazon | Responsible AI | Responsible AI team downsized. Bloomberg reported a new AI ethics monitoring team folded. | Shrunk |
| 6 Mar 2023 | Microsoft | Ethics and Society | Remaining staff told the team is eliminated. Office of Responsible AI kept. | Eliminated |
| 1 May 2023 | Microsoft | Responsible AI programme | Nearly 350 people work on responsible AI, 129 full time. | Growth disclosure |
| 18 Nov 2023 | Meta | Responsible AI | Division disbanded. Most staff move to generative AI, some to AI infrastructure. | Reassigned |
| Feb 2024 | Responsible Innovation (RESIN) | Founder Jen Gennai leaves. About 90% of roughly 30 staff move to trust and safety. | Reassigned | |
| 1 May 2024 | Microsoft | Responsible AI programme | Team grows from 350 to more than 400. More than half full time. | Growth disclosure |
| May 2024 | OpenAI | Superalignment | Disbanded after Ilya Sutskever and Jan Leike leave. Work folded into other research. | Reassigned |
| 24 Oct 2024 | OpenAI | AGI Readiness | Disbanded after Miles Brundage leaves. Members reassigned. | Reassigned |
| 31 May 2025 | Meta | Product risk review | Internal documents say up to 90% of risk assessments will be automated. | Process automated |
| Feb 2026 | OpenAI | Mission Alignment | Team of 6 or 7 disbanded. Lead Joshua Achiam named chief futurist. | Reassigned |
Last verified 13 November 2026. Dates are the event date where reporting gives one, otherwise the month of first reporting. Headcounts are as reported by the named source and are approximate where the source says so.
Reassignment is the usual move for AI ethics teams
The popular story is that tech companies fired their ethicists. The record supports that for Twitter and for part of Microsoft. For the rest of the ledger, the dominant pattern is a named team being dissolved and its people distributed into groups with other goals.
That distinction matters for anyone judging corporate AI ethics from the outside. A layoff is a clear signal. A reassignment is a claim that the work continues somewhere else, and that claim is almost impossible to verify without the company's cooperation.
Meta, November 2023: responsible AI moves into the product group
Meta disbanded its Responsible AI division, according to a Meta spokesperson quoted by CNBC on 18 November 2023. Most members moved to the Generative AI product division. Others joined the AI Infrastructure team. The spokesperson said Meta would "continue to prioritize and invest in safe and responsible AI development."
Eighteen months later the review process itself changed. NPR reported in May 2025 that internal documents showed up to 90% of Meta's product risk assessments would be automated. Meta said humans still handle "novel and complex issues." Put the two events together and the oversight function moved first into the product group, then partly into software.
Google, February 2024: responsible innovation moves into trust and safety
Google's Responsible Innovation team, known as RESIN, had about 30 people. After founder Jen Gennai left, Google split it. Spokesperson Brian Gabriel said 10% would stay in RESIN and 90% would move to trust and safety. No layoffs were reported.
Trust and safety teams protect a platform and its users from abuse. Ethics review teams ask whether a product should ship in its current form. Those are related jobs with different incentives, and the move shows which question Google chose to centre.
OpenAI, three times in 21 months
OpenAI's record is the densest. Superalignment, announced in July 2023 with a promise of 20% of the company's computing power over four years, was disbanded in May 2024 after its co-leads left. In October 2024, CNBC reported the AGI Readiness team was disbanded and its members reassigned. In February 2026, TechCrunch reported that the Mission Alignment team of 6 or 7 people was disbanded and its lead named chief futurist.
Each time, OpenAI said the work continued across the organisation. That may be true. It also means the one structure an outsider could point to, a named team with a named leader, disappeared 3 times. We think a company that repeatedly says the work is "everywhere" owes the public a way to check that it is somewhere.
Microsoft cut one AI ethics team and grew the programme
Microsoft is the only company in the ledger that appears on both sides. It is also the only one that published a headcount for its responsible AI work. That makes it the most useful case for understanding what a cut actually means.
The Ethics and Society timeline
The Ethics and Society team sat inside Microsoft's AI organisation. Its job, according to former employees quoted by Platformer on 13 March 2023, was making sure Microsoft's responsible AI principles showed up in product design. It peaked at roughly 30 people in 2020. A reorganisation in October 2022 cut it to roughly 7.
On 6 March 2023, the remaining staff were told on a call that the team was being eliminated. Platformer noted that Microsoft still maintained its Office of Responsible AI, which writes the rules. What disappeared was the group embedded with product teams. That is a specific loss, not a general one, and it fell in the same quarter Microsoft was racing OpenAI models into its products.
What 129 full-time staff tells you
Seven weeks after the last Ethics and Society cut, Microsoft's chief responsible AI officer Natasha Crampton wrote that nearly 350 people worked on responsible AI, with 129 of them full time. That is 37% of the total. The rest held responsible AI duties as part of another job.
A year later, Bloomberg reported that Microsoft's first transparency report put the team at more than 400, with more than half full time. The growth included existing employees moved into the team as well as new hires. So full-time headcount moved from 129 to something above 200. That is real growth, and it is also partly relabelling.
Our view is that Microsoft deserves credit for publishing a number at all. Five of the 6 companies in this ledger have never put a comparable figure on the record. The number can be argued with, which is exactly why it is worth more than a reassurance. How those disclosures compare in depth is covered in our scorecard of responsible AI transparency reports.
The AI ethics layoffs that came with no explanation
Two companies in the ledger cut AI ethics staff with almost no public account. In both cases, the record rests on former employees and reporters rather than the company.
Twitter laid off all but 1 member of its ML Ethics, Transparency and Accountability team in November 2022, in the first weeks after the ownership change. The team audited Twitter's algorithms for unintended harm and bias. Its director, Rumman Chowdhury, told CNBC it had 17 people. The company now operates as X. It has published no replacement structure that we could find.
Amazon's record is thinner still. CNBC reported that Amazon downsized its responsible AI team in March 2023. Bloomberg, citing people who asked not to be named, reported in April 2023 that Amazon had folded a new team monitoring the ethics of its AI efforts. Neither report gave a headcount, and Amazon did not describe the change publicly.
A company is entitled to reorganise without a press release. But a business that sells AI to other businesses, as Amazon does, should expect its customers to ask what replaced a cut oversight function. Silence is an answer, and buyers should record it as one.
What an AI ethics team actually is
"AI ethics team" covers at least three different jobs. Treating them as one is why the public debate keeps talking past itself.
The first is a central policy office that writes rules and reviews high-risk launches. Microsoft's Office of Responsible AI is the clearest example, and it survived. The second is an embedded team that works inside product groups and shapes design before launch. Microsoft's Ethics and Society team and Google's RESIN were closer to this model, and both were cut or moved. The third is a research group that publishes on risks, such as Google's Ethical AI team or OpenAI's Superalignment.
Which shape gets cut
The ledger suggests embedded and research teams are the ones that go. Central policy offices tend to survive, probably because regulation needs someone to own compliance. That pattern matches the rise of formal frameworks such as the NIST AI Risk Management Framework questions now appearing in RFPs and certification schemes like ISO 42001.
The trade is visible. A compliance office can tell you a product met a standard. An embedded team can tell you a product should have been built differently. Companies are keeping the first and shedding the second. We think that is the most important finding in this ledger, and the one least visible in a simple count of layoffs.
| Shape | Job | Examples in the ledger | Fate |
|---|---|---|---|
| Central policy office | Writes rules, reviews sensitive launches | Microsoft Office of Responsible AI | Kept, and the programme around it grew |
| Embedded product team | Shapes design inside product groups | Microsoft Ethics and Society, Google RESIN, Meta Responsible AI | Eliminated or moved into product and trust and safety |
| Research group | Publishes on risk and long-term safety | Google Ethical AI, Twitter ML ethics, OpenAI Superalignment and AGI Readiness | Leaders exited, teams cut or dissolved |
Where this AI ethics ledger is weakest
This post makes a claim about corporate behaviour from public reporting. That method has three weaknesses, and each one cuts against our framing.
Cuts are news, growth is a press release
Reporters cover a disbanded ethics team. They rarely cover a quiet hire of 20 risk reviewers. The 13 to 2 split in this ledger partly measures what gets reported, not what happens. Google, Meta and Amazon may have grown responsible AI headcount elsewhere without saying so. We could not find a published figure for any of them, which is a finding about disclosure, not proof of decline.
Reassignment is not abandonment
Every reassigning company said the work continued. Embedding ethics specialists inside product teams is a defensible design, and some practitioners argue it beats a separate team that product managers learn to route around. If that view is right, 5 of our 13 negative events are neutral or even positive. We do not think the evidence settles it either way, because none of these companies has published what the reassigned staff now do.
There are also gaps we could not close. Several headcounts are approximate, and the Google RESIN figures come from a secondary outlet because the original Wired report could not be retrieved. Anthropic, IBM, Salesforce and Apple are not in the ledger, because we found no dated team-level event for them in this pass. A wider sample could change the ratio.
Business ethics questions to ask an AI vendor
The practical use of this ledger is procurement. If you buy AI from any of these companies, or from a startup built on their models, their oversight structure is now part of your own risk. The questions below turn the ledger into something you can send. Pair them with the broader AI risk questionnaire for RFPs.
| Ask this | A useful answer | A weak answer |
|---|---|---|
| How many people work on responsible AI full time? | A number and a date, like Microsoft's 129 of about 350 | "It is everyone's responsibility" |
| Who can stop a launch on ethics grounds? | A named role with documented authority | A review committee with no veto |
| What happened to the team that did this work before? | A dated account of any reorganisation | No record, or "the work continues" |
| Which risk reviews are automated? | A stated share and which categories stay human | No answer, after Meta's 90% plan became public |
The third row is the one most buyers skip. A vendor that can describe its own reorganisations in dates is a vendor that keeps records. Employees who see a gap between policy and practice also now have legal cover in some states, covered in our guide to SB 53 whistleblower protections. How companies talk about AI and headcount more broadly is tracked in our record of AI layoff statements and their corrections.
Frequently asked questions
Which tech companies cut their AI ethics teams?
Public reporting shows Microsoft eliminated its Ethics and Society team in March 2023, Twitter laid off all but 1 member of its ML ethics team in November 2022, and Amazon downsized its responsible AI team in March 2023. Meta, Google and OpenAI dissolved named teams and reassigned staff, Meta in November 2023, Google in February 2024 and OpenAI three times between May 2024 and February 2026.
Did Microsoft lay off its AI ethics team?
Yes, one of them. Microsoft cut its Ethics and Society team from roughly 30 people in 2020 to about 7 in October 2022, then eliminated it in March 2023, Platformer reported. Microsoft kept its Office of Responsible AI and said in May 2023 that nearly 350 people worked on responsible AI, 129 full time. It reported growth to more than 400 in 2024.
Why did Meta disband its responsible AI team?
Meta did not give a detailed reason. In November 2023 a spokesperson confirmed the Responsible AI division was disbanded, with most members moving to the Generative AI product group and others to AI infrastructure. Meta said it would keep investing in safe and responsible AI. In May 2025 NPR reported Meta planned to automate up to 90% of product risk assessments.
What happened to OpenAI's superalignment team?
OpenAI disbanded its Superalignment team in May 2024 after co-leads Ilya Sutskever and Jan Leike left. It had been promised 20% of OpenAI's computing power over four years when announced in July 2023. OpenAI said the work was folded into other research. It later dissolved its AGI Readiness team in October 2024 and its Mission Alignment team in February 2026.
Is a responsible AI team the same as an AI ethics team?
Not always. Some responsible AI teams are central policy offices that write rules and handle compliance. Others are embedded in product groups and shape design, and some are research groups publishing on risk. The public ledger shows central offices mostly survived, while embedded and research teams were the ones most often cut, dissolved or reassigned between 2020 and 2026.
How can a buyer check if an AI vendor still has an ethics team?
Ask for a full-time headcount with a date, the named role that can stop a launch, and a dated account of any reorganisation of the team. Microsoft has published a headcount, so the request is reasonable. A vendor that only says the work is shared across the company has given you a statement, not evidence, and you should record it that way.
Where to start
Pick the two AI vendors you spend the most with. Send each one the four questions in the table above, word for word, and give them two weeks.
Then file the answers next to the contract. If a vendor later reorganises its oversight function, as 5 of the 6 companies here have, you will have a dated baseline to compare against instead of a press statement.
References
- Platformer, Zoë Schiffer and Casey Newton, Microsoft just laid off one of its responsible AI teams, 13 March 2023. Used for the Ethics and Society timeline and the Office of Responsible AI.
- Microsoft On the Issues, Natasha Crampton, Reflecting on our responsible AI program, 1 May 2023. Used for the 350 and 129 figures.
- Bloomberg, Microsoft boosts responsible AI team from 350 to 400 personnel, 1 May 2024, and Amazon's Twitch safety, AI ethics job cuts raise concerns, April 2023, both via Yahoo Finance syndication.
- CNBC, Tech layoffs ravage the teams that fight online misinformation and hate speech, 26 May 2023; Meta breaks up its Responsible AI team, 18 November 2023; OpenAI disbands AGI Readiness team, 24 October 2024.
- BTW Media, Google splits up a key AI ethics watchdog, 1 February 2024. Used for RESIN figures. Secondary to Wired's original report.
- TechCrunch, OpenAI disbands mission alignment team, 11 February 2026.
- NPR via OPB, Meta plans to replace humans with AI to assess privacy and societal risks, 31 May 2025.
- Engadget, Google has fired another AI ethics leader, 19 February 2021. Used for the Gebru and Mitchell exits.
Weakest point in the source base: almost every entry rests on press reporting and former employees, not company filings, and the Google RESIN figures come from a secondary outlet. Only Microsoft has published headcount. Last verified 13 November 2026.
Related reading