Is Anthropic Halal? A Sharia Analysis for Muslim Investors (2026)

Verdict: Not yet determinable for retail investors, and the activity screen raises serious concerns. Anthropic is not publicly traded, so most Muslim investors cannot buy shares. For those with access to pre-IPO shares or planning to participate in the expected autumn 2026 IPO, the core AI business is permissible, but the defence and intelligence contracts are a material problem under AAOIFI Standard 21. Until the S-1 filing reveals revenue segmentation and audited financial ratios, a definitive public-market ruling is not possible.

Anthropic filed a confidential draft S-1 with the SEC on June 1, 2026, and is expected to list in the autumn at a valuation near $965 billion. The company has raised $65 billion in its Series H round, giving it one of the largest private-market capitalisations in history. For Muslim investors, the question is not only whether AI software is halal; it is whether a company that sells that software to military and intelligence agencies can pass the business activity screen.

Anthropic PBC is an American artificial intelligence company founded in January 2021 by former OpenAI employees, including siblings Dario and Daniela Amodei. It develops the Claude family of large language models and positions itself around AI safety research. The company operates as a public benefit corporation and is headquartered in San Francisco.

The business has four segments.

The defence segment is not theoretical. In November 2024, Anthropic partnered with Palantir and Amazon Web Services to supply Claude to U.S. intelligence and defence agencies. In June 2025, it launched Claude Gov, a model explicitly built for classified missions. In July 2025, the U.S. Department of Defence awarded Anthropic a $200 million military AI contract alongside Google, OpenAI, and xAI. The Wall Street Journal reported that the U.S. military used Claude in its 2026 raid on Venezuela, and the Pentagon investigated whether Claude played a role in a subsequent strike on an Iranian girls' school.

Because the company is private, the revenue split between commercial AI and defence is not public. The S-1, once filed publicly, will be the first audited disclosure of this breakdown.

Under AAOIFI Shari'ah Standard No. 21, Financial Paper: Shares and Bonds (adopted 2004; 2015 English edition), the activity screen asks whether the company's business lines are intrinsically permissible. Software development, licensing, and cloud services are commercial activities with no inherent prohibition. The halal vs haram in Islamic finance guide covers the default permissibility of trade.

The problem is the defence segment. Claude Gov, the Palantir partnership for classified missions, and the DoD contracts are designed for military and intelligence operations. AAOIFI Standard 21 does not name defence contracting among its prohibited activities. Whether AI systems built for combat, targeting, surveillance, or espionage constitute a prohibited business line is a judgement some contemporary screeners make by analogy to the harm-based prohibitions; it is not stated in the standard itself, and scholars differ on the permissibility of supplying lawful state defence.

Anthropic's resistance to unrestricted military use is notable. In February 2026, the company refused a Pentagon demand to remove contractual restrictions prohibiting domestic surveillance and fully autonomous weapons. It challenged the DoD's supply chain ban in court. Some scholars might view this resistance as a mitigating factor. Others would say that any defence contract at all, even with restrictions, fails the activity screen.

As a private company, Anthropic's debt structure is not fully disclosed. The $65 billion Series H round, its multibillion-dollar cloud computing commitments to Amazon Web Services and Google Cloud, and the $200 million Snowflake partnership imply significant capital movement, but the form of that capital matters. Venture funding at the Series H stage is typically equity, not debt. Cloud computing contracts are operational, not financing. Without an audited balance sheet, the standard AAOIFI financial ratios cannot be calculated.

Allah says in Surah Al-Baqarah, verse 275:

وَأَحَلَّ اللَّهُ الْبَيْعَ وَحَرَّمَ الرِّبَا

"But Allah has permitted trade and has forbidden interest [riba]." (Saheeh International, Qur'an 2:275)

When the S-1 becomes public, the three ratios can be applied: debt to market capitalisation, cash and interest-bearing securities to market capitalisation, and interest income to revenue. Until then, the riba screen is unverifiable for private shares, and the IPO price will reset all three ratios anyway.

Pre-IPO investment carries extreme gharar. There is no public market price, no dividend history, no guaranteed liquidity event, and no audited public financial statement. The confidential filing suggests an autumn 2026 IPO, but the timing, pricing, and even completion are uncertain. This makes any direct private investment in Anthropic high in uncertainty.

Once listed, the shares themselves will be well-defined and deliverable, so contractual gharar falls away. A $965 billion valuation without published profits is severe market risk, which Shari'ah permits the investor to bear, but it is prudent to distinguish that risk from the contractual gharar of the pre-IPO stage. The ethical investing framework addresses this category of valuation risk.

Anthropic does not operate gambling, betting, or games of chance. The maysir screen is not engaged by the business model itself. The speculative nature of investing in a pre-IPO AI company at a trillion-dollar valuation without profitability is a financial risk, not maysir under the AAOIFI framework, provided the investor is purchasing equity in a productive enterprise. Risk in a productive enterprise, even extreme valuation risk, is not gambling.

None of the three scholars named below has issued an Anthropic-specific ruling. What follows applies their published methodologies to the facts; the extrapolations are ours.

The primary scholarly disagreement is not about the permissibility of AI software, which is largely uncontested, but about how to treat defence contracts and the lack of public financial data.

Sheikh Joe Bradford's equity screening guidelines (2024) apply the AAOIFI standard to publicly traded stocks. For private companies, his framework still applies the activity screen. He distinguishes permissible infrastructure from prohibited weapons development. The Claude Gov model, designed for classified national security work, would likely fall on the non-permissible side of this line. His position would probably yield a non-compliant verdict once the company is public, unless defence revenue is immaterial.

Mufti Faraz Adam's stock screening methodology (2023), published through Amanah Advisors, takes a granular approach to revenue streams. His framework would require knowing what percentage of Anthropic's revenue comes from defence versus commercial AI. Because that breakdown is not public, his methodology cannot be fully applied. His general principle is that a company with a material prohibited revenue stream is non-compliant.

Mufti Taqi Usmani's equity conditions, set out in An Introduction to Islamic Finance (1998), require that the company's main business be permissible. He has not published a position on AI or military contracting; an investor extending his stricter method would ask whether products built for combat or intelligence can sit within a permissible main business, but that extension is ours, not his.

On the AI safety question, some scholars argue that Anthropic's refusal to allow unrestricted military use, and its legal challenge to the Pentagon, makes it more permissible than peers like OpenAI or xAI that accepted unrestricted DoD contracts. Others argue that the presence of any defence contract at all is the determinative factor, and corporate resistance does not cleanse the revenue.

Even if the financial ratios pass when the S-1 is filed, two activities are independently non-permissible under the dominant scholarly view.

First, the sale of AI systems for classified military and intelligence operations. The Palantir partnership, the Claude Gov model, and the DoD contracts are not incidental consulting work; they are products built for a prohibited purpose.

Second, the lack of transparency itself is not a fiqh violation, but it prevents the investor from applying the AAOIFI screen. Investing without the ability to verify revenue sources or financial ratios is a practical barrier to compliance.

For Muslim investors, the guidance depends on access.

If you do not have access to pre-IPO shares, the question is theoretical until the IPO. Wait for the S-1 filing. The S-1 will contain the first audited revenue segmentation, debt figures, and business description. Only then can the AAOIFI screen be applied rigorously.

If you hold Anthropic indirectly through a venture fund, index fund, or ETF, you do not control the underlying selection. Purification methodologies differ. Under the dividend-based approach there is currently nothing to purify because Anthropic pays no dividend. Under the AAOIFI income-based approach, purification tracks the company's impermissible earnings regardless of distribution, which cannot be calculated until the financials are public. The stock itself, if held indirectly, is a matter for the fund's Shari'ah board.

If you are considering the IPO, evaluate the S-1 carefully. Look for the revenue split between commercial AI and government contracts. Look for the debt ratio. Look for interest income. If defence revenue is material, or if the activity description confirms the continuation of the Palantir and DoD partnerships, the dominant scholarly position will likely classify the stock as non-compliant.

You can run any stock through the AAOIFI screening framework on the stock screener and screen digital assets on the crypto screener. For a comparison of how defence exposure affects compliance, see the analysis of SpaceX stock, which also fails the activity screen on military contracts.

Anthropic's core business, building and selling AI software, is a permissible commercial activity. Under AAOIFI Standard 21, the decisive questions will be whether defence work forms a material business line and whether impermissible income stays under the 5 percent threshold. Scholars differ on whether defence contracting is prohibited at all; investors following the stricter view would avoid the stock regardless of revenue share, while those following the standard screen would test materiality once the S-1 is public. Because Anthropic is not yet public, the exact financial ratios and revenue segmentation are unknown. The confidential S-1 filing means audited data exists but is not available. Muslim investors should wait for the public S-1 before making a determination. If the defence segment is material, the dominant scholarly position will classify the stock as non-compliant. If the company separates its commercial AI from its government contracts through a structural spin-off, that standalone commercial entity would be a more plausible compliant candidate. Until then, the answer is: wait, and watch the filing.

Can I buy Anthropic stock now? No. Anthropic is privately held. The company filed a confidential draft S-1 on June 1, 2026, and an IPO is expected in autumn 2026. Only accredited investors with access to private markets can hold shares today.

Is AI software itself halal or haram? AI software development and licensing are generally permissible commercial activities. There is no inherent prohibition in creating or selling software under classical fiqh or AAOIFI standards.

Do Anthropic's defence contracts make it haram? It depends on the position followed. Some contemporary screeners treat military and intelligence AI as a prohibited activity; others treat lawful defence supply as permissible or as a purification matter if immaterial. The question is genuinely contested. Some scholars view Anthropic's resistance to unrestricted military use as a mitigating factor, but the presence of the contracts themselves is the primary concern.

What should I wait for before deciding? The public S-1 filing. It will contain audited revenue segmentation, debt figures, cash holdings, and a description of all business lines. Only then can the AAOIFI financial and activity screens be applied rigorously.

Does Anthropic pay a dividend? No. Anthropic has never paid a dividend. This means there is no purification obligation for current holders, but it does not change the permissibility of the stock itself.

How does Anthropic compare to SpaceX or Microsoft on defence? All three have defence exposure. SpaceX fails the activity screen on Starshield and content. Microsoft's defence contracts are a smaller share of revenue and are often infrastructure rather than targeting. Anthropic's Claude Gov is explicitly designed for classified national security work, which places it closer to the non-permissible side. See the SpaceX analysis for a detailed breakdown.

Where can I verify the data when the S-1 is filed? The stock screener on this website updates when new SEC filings arrive. The figures above are based on public reporting from Reuters, The Wall Street Journal, and Anthropic's own newsroom as of June 2026.