From industry darling to pariah: What's going on with Anthropic?
Anthropic, once the industry's darling, is facing a wave of criticism over its stance on open-source AI, aggressive scare tactics, and competitive practices that have alienated partners.
Image processing: Geektime, AI
For a long time, Anthropic, the AI company founded by seven OpenAI alumni, including brothers Dario and Daniela Amodei, was considered the industry's darling. Praise for Claude Code was constant, its enterprise market share soared, industry influencers documented their switch from ChatGPT to Claude, and its models created significant buzz. Within the industry, it earned a reputation as a successful firm that maintains its principles—a key advantage in the fight for Silicon Valley talent. However, lately, something has changed, and not for the better, at least in the relationship between the company and other tech firms.
An initiative everyone loves, except one
One of the biggest topics of conversation in the industry in recent weeks is the use of open AI models and the need to protect and promote them. Quite a few initiatives have emerged, including one promoted by NVIDIA, the company whose hardware drives the entire industry. NVIDIA's initiative, designed to ensure that AI models do not remain solely closed and proprietary, was quickly signed by nearly 70 companies, including Meta, Microsoft, OpenAI, Google, and even Elon Musk's SpaceX (which includes the AI activity of xAI). But one big name was missing: Anthropic.
Instead of joining the initiative, the company published a response in the name of CEO and founder Dario Amodei: "Anthropic has never supported blocking the release of open-source models," Amodei wrote. According to him, the company's lack of support for NVIDIA's initiative and the fact that it does not publish open-source models stem from the fear that such models would give authoritarian regimes the ability to compete with the USA and "achieve military superiority or oppress their people."
Anthropic's stance led to sharp criticism. "The entire tech industry (except Anthropic) has expressed its support for open-source AI," said David Sacks, an investor and former AI advisor to US President Donald Trump, adding: "They won't stop until they destroy the open-source AI field. The rest of the players in the industry should be wary of them." Peter Steinberger, an OpenClaw developer now at OpenAI, wrote: "Proud of OpenAI for the support. Anthropic is keeping the right to remain silent." Investor Bill Gurley of the Benchmark fund added that he is "excited to see Anthropic admit that the only problem with open models is that they compete with their business strategy."
Competitors who didn't expect it
Criticism didn't stop there. An article in the Wall Street Journal reveals that companies that did not expect to compete with Anthropic, and even worked closely with it, were surprised to find it suddenly entering their territory. Design giant Figma was surprised to find that Anthropic—until then considered a business partner—was going for its throat with Claude Design, released last April.
Dylan Field, CEO of Figma, said that Anthropic "was not entirely honest in the discourse" between the companies, meaning it did not signal in advance that it was developing a product to compete with Figma. Sarah Sacks, head of AI at Notion, said: "The industry has learned a lot from the dynamics around this case. Some companies have apparently realized that they trusted too much."
Scare tactics that everyone is tired of
Another area where Anthropic receives a flood of criticism is its alleged scare tactics. It started with CEO Dario Amodei's statements that AI would eliminate most white-collar jobs—a statement he later walked back—and the series of events around Mythos, the model unveiled in April with warnings about its cyber capabilities. Since then, the model has leaked, reached the public with aggressive defense mechanisms, and been blocked by the US government until its re-release. Many claimed the fear-mongering about Mythos or Fable was mainly a marketing trick.
Microsoft CEO Satya Nadella, in a leaked conversation with employees, said: "When you use Fable and it blocks you for every little thing, it makes you wonder: when was the last time you saw a creation tool that tries so hard to censor and manage you? It just doesn't make sense." Academia has also criticized the company; a senior official at Columbia University's School of Engineering stated that while they develop excellent products, they "create fear in the general population around these models, which gives them power over regulators so that they act in their favor."
An example of this influence is reports that the administration of President Donald Trump is considering significant sanctions against Chinese AI companies on the grounds that they are "distilling" models like Claude and GPT. Furthermore, Anthropic was forced to pay $1.5 billion in damages for copyright violations while training its models on pirated books.
At this stage, Anthropic shows no signs of backing down. As a powerful player with an estimated value nearing a trillion dollars, it is doubtful it will easily give up on issues it perceives as existential.
