Artificial Intelligencer-Anthropic walks a tightrope between courtroom alarm and investor calm

By Kenrick Cai

March 11 (Reuters) - (Artificial Intelligencer is published every Wednesday. Think your friend or colleague should know about us? Forward this newsletter to them. They can also subscribe here or email me to share any thoughts.)

Meta META.O made a curious acquisition this week in buying Moltbook, a Reddit-like site where AI agents interact with one another.

More than likely, the deal is less about the Facebook parent company integrating Moltbook into its social media platforms than about integrating the technology that underlies it.

Moltbook is built on OpenClaw, an open-source AI tool that can perform tasks such as booking flights and organizing email. OpenAI hired the creator of OpenClaw in February as part of its own push to develop AI agents.

OpenClaw, which launched in November, almost instantly became one of the fastest-growing projects in the history of popular developer platform GitHub. By now, it has escaped the Silicon Valley bubble. The rapper Baby Keem, known for his collaborations with his cousin Kendrick Lamar, diverted from his regular programming on social media platform X to post in February, “how do u fix openclaw internal reasoning leaking,” poking at the security vulnerabilities associated with the tool.

In China, government officials in several cities are pushing to establish an industry revolving around OpenClaw, as my colleague Eduardo Baptista reported earlier this week. The tool has already spawned a number of "one-person companies."

All the OpenClaw-related moves provide a preview of what could be the next battleground in AI: personalized agents that can execute real-world tasks.

In this week’s issue, we look at the delicate balance Anthropic must strike in describing its business outlook as it takes the Pentagon to court — and where around the world ChatGPT is not the dominant chatbot. Scroll on.

OUR LATEST REPORTING IN TECH AND AI:

• Anthropic has strong case against Pentagon blacklisting, legal experts say

• Meta unveils plans for batch of in-house AI chips

• AI startup Thinking Machines clinches capital and a major chip supply deal from Nvidia

• Oracle rallies as strong revenue forecast eases concerns over massive AI bets

• China pins hopes on society-wide AI push to add jobs, rejuvenate economy

• US mulls new rules for AI chip exports, including requiring US investments by foreign firms

ANTHROPIC’S DELICATE DANCE

Anthropic’s lawsuit filed Monday to block the Pentagon’s restrictions on its technology will force the AI lab to tread a fine line in presenting the potential business impact of its placement on a national security blacklist as both perilous and secure.

The lawsuit is the latest escalation of a monthslong standoff over Anthropic’s refusal to remove guardrails against using its AI for autonomous weapons or domestic surveillance. President Donald Trump has directed federal agencies to cease using Anthropic technology, and the Pentagon has designated the company as a supply chain risk.

To a federal judge, Anthropic is arguing that billions of dollars in revenue in 2026 alone could be in jeopardy.

Already, it lost a contract for a federal deployment of its Claude generative AI model that was expected to bring in more than $100 million, and negotiations with financial institutions worth $180 million have been disrupted, according to Chief Commercial Officer Paul Smith in federal court filings. Hundreds of millions more in work carried out for the Defense Department are at immediate risk, per CFO Krishna Rao.

Rao also said that the government blacklist could undermine investor confidence and require Anthropic to increase fundraising costs.

And so Anthropic, which is reportedly exploring an IPO as early as this year, must concurrently assure investors that its business outlook remains healthy.

In a statement last week, CEO Dario Amodei said that the restrictions have “a narrow scope” and enterprises may continue to use its technology for work unrelated to the Pentagon.

Anthropic’s ability to maintain revenue momentum — annualized revenue is now $19 billion, up from $14 billion in February — will depend on its ability to retain and attract enterprise customers, which account for some 80% of its business.

One Anthropic backer told me that, anecdotally, the increased publicity has so far drawn more business customers over to Anthropic. But Smith stated in the court filing that more than 100 enterprise customers have already expressed “deep fear” about the repercussions of working with the company.

Meanwhile, the publicity has provided a tailwind of consumer users who lifted Anthropic’s Claude chatbot to the top of Apple’s App Store.

Anthropic is not in a rush to raise capital. The company just closed a $30 billion fundraise last month. Two investors in that funding round told me they feel safer about Anthropic’s cash burn rate, which is helped by its enterprise-heavy business that doesn’t require as many computing costs to service non-paying users, as in the case of OpenAI’s ChatGPT.

Beyond military contracts, the impact on business remains to be seen. One Anthropic investor told me that they are unconcerned about a revenue slowdown. While investors would generally be leery of a company being embroiled in a lawsuit with the federal government, Anthropic remains too attractive an IPO candidate for the current situation to dramatically shift IPO plans, this person said.

CHART OF THE WEEK

As Anthropic vies to maintain its enterprise stronghold, OpenAI’s ChatGPT continues to reign supreme among consumers. Zoom in geographically, though, and you’ll notice challengers making inroads, as this data from venture capital firm Andreessen Horowitz’s latest AI report indicates.

Notably, the Chinese market is dominated by homegrown chatbots, the result of restrictions on foreign AI tools that do not comply with state policy. Meanwhile, Western sanctions on Russia have allowed a range of alternatives to break through, led by China’s DeepSeek. Google’s Gemini appears to be the only close competitor to ChatGPT in other markets, with a tighter gap in South Korea and Japan.