Stability AI Cuts 10% of Staff Following CEO’s Departure

One day after announcing the accessibility of Stable Diffusion 3 to certain developers, via an API and a new content creation platform, the start-up Stability AI indicated on April 18 that it would separate itself from more than 10% of its workforce. All UK staff “non-essential”i.e. the majority of staff working in the United Kingdom, are affected, according to a source cited by Sifted.

“Reduce the cost base”

Advertisement

According to an internal memo, the dismissal of 20 employees would be justified for “resize” the company, after a period of unsustainable growth. A decision forming part of a “strategic plan to reduce (the) cost base, strengthen support from investors and partners and enable teams to continue to grow”, said Shan Shan Wong and Christian Laforte, new interim co-CEOs of Stability AI, in an email sent to employees Wednesday evening. The dismissed employees were individually notified.

These layoffs come after numerous movements within the start-up. At the end of March, co-founder Emad Mostaque resigned as CEO. He then declared on X (formerly Twitter): “The concentration of power in AI is bad for all of us. I have decided to resign to resolve this issue at Stability AI and elsewhere.” Shan Shan Wong, then COO, and Christian Laforte, CTO, were named interim co-CEOs. The young company is still looking for a new permanent CEO.

Stability AI is experiencing a lot of turbulence

Also in March, Robin Rombach, one of the original developers of the Stable Diffusion model, also left the company, and the former CTO, Thomas Mason, was replaced. Last year, around ten senior executives of the start-up, occupying product design or engineering positions, slammed the door or were fired. A period during which the start-up also failed to raise funds, bringing its valuation to four billion dollars.

Advertisement

However, the London start-up has been increasing its announcements since the start of the year. At the end of February, it launched Stable Diffusion 3, an image generation model intended to compete with OpenAI's Sora and Google's Gemini. In mid-January, it unveiled Stable Code 3B, allowing developers to complete their code with a level of relevance comparable to models twice the size.

Selected for you

Advertisement