As Microsoft, Meta, Google, Amazon, Twitter, and others cut their responsible AI teams, experts worry about potential abuses, disinformation, and hallucination
The responsible-AI function these firms built is being dismantled just as deployment accelerates. The same companies now cutting those teams were founders of the cross-industry AI ethics group formed in 2016 to study AI's impact on jobs and warfare, and the cuts land weeks after experts called Microsoft's new Bing experiment dangerous to the public. Internal reporting had already flagged the mechanism: pressure inside Meta and Google to move faster amid ChatGPT attention was sweeping safety concerns aside months before the teams themselves were cut.
What makes the timing consequential is that the risks experts name — hallucination, disinformation, exaggerated capability claims — are not hypothetical. Researchers warned in 2022 that businesses' hype was already distorting policymakers' views of AI's power and fallibility, meaning the guardrail staff leaving are the same people who would have corrected the record.
First-order effects
Responsible-AI researchers at Microsoft, Meta, Google, Amazon, and Twitter lose their institutional platform overnight, removing the internal review layer right as Bing-style consumer deployments expand.
Second-order effects
With in-house ethics teams thinned, technical fixes fill the vacuum downstream — a path Google, Amazon, Cohere, and Mistral later take by intensifying hallucination-reduction work on data quality and fact-checking rather than governance.
Third-order effects
If safety review keeps losing to release speed, accountability shifts from company-embedded teams to external actors — regulators, researchers, and the cross-industry ethics bodies the firms themselves created in 2016 but no longer staff internally.
The trend: Frontier AI firms are trading embedded responsible-AI capacity for deployment speed, shifting the burden of oversight from inside the labs to post-hoc technical patches and outside scrutiny.
@drgavinm ... True - and I would argue that most of these firms have over-hired in the last few years hence the layoffs. Have they overhired in this very specific and growing space? I highly doubt it.
Caring about AI ethics was a zero interest rate phenomenon. While you're in meetings debating how many AI angels can dance on the head of a pin, 11 people at midjourney and 300+ at OpenAI are disrupting your entire industry. You snooze, you lose. https://www.ft.com/...
@sarahdrinkwater ... We know nothing other than people were let go in relatively small numbers across a number of huge companies. We know nothing about their roles. We know nothing about the proportion of cuts in these teams compared with others. It's an absolute nothing story.
@mr_james_c ... Respectfully - spent time speaking to these teams/ funding adjacent work at Omidyar network. A few are personal friends. And ofc still speak to many big tech folks as an ex big tech person. Much we don't know but it still feels like a story; esp in context of toda…
This is my biggest concern right now, when the responses are completely wrong but most people who just think the responses are correct. https://twitter.com/...
(It may, perhaps, be possible to believe both that there are important questions people building products need to think about, and also think that not everyone on all of these teams was necessarily doing much of practical value towards that aim...
Especially given some of these companies doubled or tripled their total headcount in the last decade. As a joke went, “how many people work at Google? Oh, about half")
["The numbers of staff affected remain in the dozens and represent a small fraction of the tens of thousands of tech workers whose jobs were cut by companies in response to a broader industry downturn."] So why is this a story then?
@mr_james_c ... You're right. The story should be why the numbers involved were already so comparatively low for what should be an important issue that needs to be taken more seriously by providers.
AI ethicists are good. You know what's better? Encouraging the developer and the product manager to think of the real-world impact of their algorithmic recommendation system. Not just the bottom-line. Ethics is not a feature to be added. https://www.ft.com/... https://twitter.com…
AI ethics teams have been cut by Meta, Twitch, Microsoft, Google and Twitter while companies release new artificial intelligence products. “It is problematic when responsible AI practices are deprioritised for competition or for a push to market” https://www.ft.com/...
It is shocking how many members of responsible #AI are being let go at a time when arguably,you need more of those teams than ever Tech giants have been slashing staff dedicated to evaluating ethical issues around deploying AI,raising worries about safety https://www.ft.com/...