As of 2022 Anthropic pursues its mission primarily through generative model research and AI safety research.

Judge for 2022. TRUE: Anthropic's 2022 public materials and its activity (published papers, no commercial product launched) show its work was mainly research on generative models (such as large language models) and AI safety research. MOSTLY TRUE: research was the main activity, though the materials frame it a little differently (for example as alignment or interpretability research) or there was minor early product or partner testing. MIXED: research and commercial deployment were roughly equal priorities in 2022. MOSTLY FALSE: Anthropic was already mainly pursuing commercial deployment in 2022, with research secondary. FALSE: its work in 2022 was not research of this kind.

Set before the research began.

1. E1 Anthropic's quest for better, more explainable AI attracts $580M | TechCrunch TechCrunch · 29 April 2022

   “The company has already published several papers looking into, for example, reverse engineering the behavior of language models to understand why and how they produce the results they do.”

2. E2 Anthropic raises Series B to build robust AI Anthropic · 29 April 2022 Own record

   “Since its founding at the beginning of 2021, Anthropic has conducted research into making systems that are more steerable, robust, and interpretable.”

3. E3 Anthropic raises Series B to build robust AI Anthropic · 29 April 2022 Own record

   “We’ve made strong initial progress on understanding and steering the behavior of AI systems, and are gradually assembling the pieces needed to make usable, integrated AI systems that benefit society.”

4. E4 Constitutional AI: Harmlessness from AI feedback Anthropic · 15 December 2022 Own record

   “We experiment with methods for training a harmless AI assistant through self-improvement, without any human labels identifying harmful outputs.”

TechCrunch reported in April 2022 that Anthropic had already published several papers, including work on reverse engineering how language models behave (E1). Anthropic's April 2022 Series B announcement describes its work since founding as research into making systems more steerable, robust and interpretable. It also says the company is gradually assembling the pieces for usable, integrated systems (E2, E3). Its December 2022 Constitutional AI paper is a further example of language-model safety research (E4). Together these show research was the main activity, though it was framed mostly as interpretability and steerability research. None of the evidence directly confirms that no commercial product had launched.