Amazon CEO Andy Jassy may have been the source of security concerns that led Anthropic to cut off worldwide access to two models on Friday.

The Wall Street Journal reports that Jassy told Treasury Secretary Scott Bessent and other government officials that Amazon researchers used Anthropic’s Claude Fable 5 to obtain information that could be used in cyberattacks. The government subsequently imposed an export control ban on the Fable 5 and Mythos 5 models.

An Amazon spokesperson said in a statement that while it’s “not uncommon for governments to seek our counsel on potential security risks,” the company does not “share the details of those discussions.”

The spokesperson also pointed to an update stating that AWS has been affected by the model cut off.

The Information and Reuters similarly reported that Amazon (a major Anthropic investor) had communicated concerns about the security of Anthropic’s models.

David Sacks, Trump’s former AI czar who now co-chairs the President’s Council of Advisors on Science and Technology, offered his own account of the discussions, claiming that “a highly credible trusted partner of both Anthropic and the USG […] came forward with a jailbreak.”

Sacks added, “The Admin asked [Anthropic CEO Dario Amodei] to fix the jailbreak or de-deploy the model. Dario refused.”

Anthropic said in a blog post that the capabilities apparently causing government concern are already available in other publicly accessible models.

This post has been updated with a statement from an Amazon spokesperson.

Anthropic’s sudden move to suspend access to its newest AI models following a U.S. government directive has raised fresh questions across the global technology industry. In India, the decision has reignited a long-running debate over whether one of the world’s largest AI markets can afford to rely on technologies built and controlled elsewhere.

The announcement came late Friday, when Anthropic said it had received the U.S. government directive requiring it to suspend access to its recently launched Fable 5 and Mythos 5 models for all foreign nationals, including its own foreign national employees. The move came shortly after the company announced a partnership with Indian IT services giant Tata Consultancy Services to expand enterprise AI adoption in India, underlining how closely the country’s AI ambitions have become tied to technologies developed and governed in the U.S.

While the broader implications remain unclear, some reports said the initial security concerns were first reported to the government by Amazon CEO Andy Jassy. And The Information said the White House is unlikely to extend similar restrictions to other AI companies and is privately blaming Anthropic’s handling of alleged jailbreak vulnerabilities. Anthropic has disputed the government’s characterization and argued the action should not have been taken.

Regardless, the development has triggered debate among Indian founders, investors, and policy experts over whether the country should accelerate efforts to build domestic AI capabilities, deepen investment in open-source alternatives, or continue relying on a handful of U.S. frontier model providers. For some, the episode is a wake-up call on technological dependence. For others, it is a reminder that access to increasingly critical AI systems can be shaped by geopolitical decisions beyond India’s control.

India has become one of the most important markets for frontier AI companies. Anthropic and OpenAI have both described the South Asian nation as their second-largest market after the U.S., reflecting its growing importance in the global AI race. The companies have already set up their offices in India, expanded local hiring, partnerships, and enterprise initiatives in recent months, betting on India’s vast base of developers, startups, and businesses to accelerate adoption of their latest technologies.

For many in India’s technology sector, Anthropic’s Friday announcement was about more than just one AI company. It reopened questions about the country’s long-term AI strategy and whether India could afford to remain dependent on a small number of foreign frontier AI providers.

“It completely changes things,” said Aakrit Vaish, founder of Indian AI venture platform Activate, referring to Anthropic’s decision. “I think this materially changes the way all of us should be thinking about sovereign AI in India.”

Vaish told TechCrunch that he woke up on Saturday morning “shocked and confused” by the announcement and said it strengthened the case for developing domestic AI capabilities. He expects startups to increasingly turn to open-source models and plans to encourage companies in his portfolio to reduce their dependence on a small number of frontier AI providers.

For some founders, the bigger concern was what restrictions on frontier AI access could mean for competitiveness. Vijay Rayapati, co-founder and CEO of Atomicwork, told TechCrunch that the episode highlighted the risks facing startups whose teams span multiple countries if access to advanced AI systems increasingly becomes subject to geopolitical restrictions.

Atomicwork has around 25 employees in the U.S., though much of its product engineering team is based in Bengaluru, India.

“If your AI team is not made up entirely of U.S. citizens, you are at a competitive disadvantage,” Rayapati said, arguing that unequal access to frontier AI models could give some companies a significant edge over rivals.

The concern comes as parts of India’s tech sector are already grappling with questions about how AI could reshape the economics of global talent. This week, U.S. real estate technology company Opendoor shut its India office less than two years after expanding in the country, with CEO Kaz Nejatian citing a push to bring operational work closer to customers in the U.S. and a shift toward smaller AI-native teams.

While Opendoor did not specify how much of the decision was driven by AI-related efficiencies, the move added to a broader debate about how advances in AI could affect the future of global technology work and what that might mean for India’s position as an engineering talent hub.

Beyond Anthropic

In addition to startups and AI builders, the Anthropic episode also prompted a broader debate among India’s technology leaders about dependence on foreign AI infrastructure.

Sridhar Vembu, founder of Indian SaaS company Zoho, said the move showed that “technology is the ultimate weapon” and urged Indian organizations to increasingly embrace smaller and open-source models.

“What can our government do right now? Ensure that orgs in India embrace smaller models, both Indian and Chinese open source ones,” Vembu wrote on X.

Investor and former Infosys executive Mohandas Pai responded to Vembu on X, arguing that the development highlighted the need for a far more ambitious national AI strategy and calling on the government to substantially increase investments in AI, computing infrastructure, and deep technology.

“We are way behind and need a national mission to get going quickly,” Pai wrote, urging the government to create an annual ₹500 billion (about $5 billion) fund for AI and deep tech, alongside a ₹2 trillion (around $21 billion) credit guarantee program to support cloud infrastructure, hardware, and semiconductor development.

Pai’s proposal would dwarf India’s existing AI efforts. In 2024, New Delhi approved the IndiaAI Mission with an outlay of ₹103.72 billion (about $1.2 billion) over five years, aimed at expanding compute infrastructure, supporting startups, and developing indigenous AI capabilities.

Despite growing interest in AI and New Delhi’s push to develop domestic capabilities, India remains a relatively small player in frontier model development. Only a handful of startups are pursuing foundational AI models, including Sarvam, which released open-source models earlier this year. However, another high-profile AI startup, Krutrim, pivoted toward cloud and AI infrastructure services after initially positioning itself around foundational model development.

Much of India’s AI ecosystem has instead concentrated on applications and specialized models built on top of existing foundation models. Recent examples include Avataar AI, which launched a video-generation model earlier this week aimed at providing a lower-cost alternative to offerings from rivals including Google’s Veo, Kling, Luma, and Runway.

Not everyone agrees that the primary challenge is a lack of capital. Responding to Pai’s comments, Lightspeed partner Hemant Mohapatra argued that the biggest constraints to building globally competitive AI companies are talent, access to computing resources, and execution, rather than simply the size of investment commitments.

Mohapatra estimated that training a frontier AI model could cost anywhere from hundreds of millions to several billion dollars, depending on the approach, but said successful AI companies have historically scaled their capital requirements over time as adoption grew.

Yet for some policy observers, the implications extend well beyond AI startups or model providers.

Prasanto Roy, a New Delhi-based technology policy expert who advises multinational companies, said the episode would likely reinforce concerns within the Indian government about strategic autonomy, comparing it to the lesson many countries drew from Russia’s loss of access to SWIFT and other parts of the global financial system following its invasion of Ukraine.

He told TechCrunch that the move was likely to provoke a significant nationalist backlash in India and described it as a poorly considered decision by Washington, with consequences extending far beyond Anthropic itself.

“Even if this is corrected or reversed, the Anthropic episode shows there’s no such thing as a geopolitically neutral foreign LLM,” Roy said. “American AI models are bound to American geopolitics.”

Anthropic has partnered with Indian IT services giant Tata Consultancy Services (TCS) in a bid to accelerate adoption of its artificial intelligence models at enterprises.

The partnership will see TCS creating a business unit focused on deploying Anthropic’s AI models to its customers. TCS will also gain early access to new model releases, which it says it will use to build expertise, and it will provide Anthropic’s Claude AI assistant to its employee base of more than 50,000 people.

The companies said they would develop solutions for sectors like financial services, healthcare, telecommunications, and aviation.

Frontier AI companies have been securing enterprise distribution channels by partnering with firms like TCS in India. Earlier this year, Anthropic teamed up with Infosys, and OpenAI roped in Infosys and HCLTech to do something similar.

Beyond enterprise deployments, the partnership extends to several TCS businesses and platforms. Diligenta, TCS’s U.K.-based life and pensions business with over 22 million customers, plans to use Claude for customer service and process automation. Similarly, TCS iON, the company’s digital learning platform, will offer training and certification programs on Anthropic’s models.

TCS said it would contribute capabilities to Anthropic’s Claude Code ecosystem, including tools for claims adjudication and lending advisory.

Anthropic has been working to expand its footprint in India, which the company has described as its second-largest market. Over the past year, the startup has opened an office in the country, hired for leadership roles, and expanded ties with major IT services firms.

The deal comes as investors and tech companies alike have begun doubting the viability of India’s $315 billion IT services amidst the rise of AI. Shares of TCS and Infosys have fallen about 34% and 31%, respectively, so far this year.

Anthropic is donating $150 million to place 1,000 AI fellows inside nonprofit organisations across the United States. The programme, called Claude Corps, will pay early-career workers $85,000 plus benefits for a year-long placement where they help nonprofits use Claude more effectively. Applications opened Wednesday and close on July 17.

No college degree is required. Applicants must be 18 or older, hold US work authorisation, and have no more than two years of full-time work experience. The first cohort of 100 fellows starts in October 2026. Subsequent cohorts begin in January and August 2027.

Each of the 400+ host organisations will receive a $10,000 grant and free Claude credits. Anthropic partnered with CodePath, a San Francisco nonprofit that helps first-generation and low-income students enter the tech workforce, to manage recruitment and training.

“We hope this program will expand and become a pillar of our strategy to help humankind realize the benefits of AI while also managing its risks,” said Anthropic President Daniela Amodei.

The programme is modelled loosely on service corps like AmeriCorps and Teach For America, but with a corporate sponsor and a product at its centre. Fellows are trained specifically on Claude. The organisations they serve will build their workflows around Claude. When the fellowship ends, the nonprofits are left with AI infrastructure tied to Anthropic’s ecosystem.

That dual purpose has drawn criticism. Fortune noted the “fox guarding the henhouse” dynamic: a $965 billion AI company is training the nonprofit sector to depend on its own product, funded by a donation that represents less than 0.02% of its valuation. Anthropic frames it as philanthropy. Sceptics see distribution strategy wrapped in a public benefit narrative.

Regardless of the framing, the programme addresses a real gap. Most nonprofits lack the staff, budget, and technical knowledge to adopt AI tools, even when those tools could meaningfully improve operations. Anthropic’s $100M Claude Partner Network, launched earlier, targets enterprises. Claude Corps targets the organisations that cannot afford enterprise partnerships.

The timing is deliberate. Anthropic is preparing for an IPO and positioning itself as the responsible AI company in a field dominated by OpenAI’s commercial aggression and Google’s scale. A $150 million nonprofit fellowship is a narrative play as much as a product play. Whether 1,000 fellows can make a meaningful difference across 400 organisations depends on whether the programme outlasts its PR value. Anthropic’s policy framework, published this week, calls for AI’s benefits to be “broadly shared.” Claude Corps is its first concrete attempt to deliver on that promise.

Notion’s integration with Anthropic apparently had a hiccup this weekend.

Early Sunday morning, the company posted, “Anthropic’s Opus 4.7 and 4.8 models are experiencing degraded performance, which is causing a higher rate of failures for users selecting these models in Notion AI.”

As a result, Notion said it was disabling use of “all Anthropic models” in its automated productivity tool.

Twelve hours later, Notion’s head of product Max Schoening wrote that he was “astonished” at “the amount of people RT-ing this because they want a story around model quality to be the reason.” (According to the public stats on X, Notion’s post has been reposted around 1,200 times.)

“The degraded performance was a temporary service disruption,” Schoening said. “This happens. It happens to Notion, GitHub, AWS, your OpenClaw, and everything in between.”

He added that Notion has restored access to Anthropic’s models.

Meanwhile, an Anthropic spokesperson said in a statement, “A brief infrastructure issue caused elevated errors on multiple Claude models for a short period of time. The issue has since been resolved. We’re grateful to our users for their patience while we worked to restore service.”

Anthropic is expanding Project Glasswing, its joint industry initiative to find and fix critical software vulnerabilities using AI, to about 150 new organizations across more than 15 countries, the company said Tuesday. 

The news comes a day after Anthropic said it had filed confidentially for an initial public offering, following a $65 billion funding round at a nearly $1 trillion valuation. 

Anthropic’s Claude Mythos is at the heart of Project Glasswing. The AI firm dubbed the model its most powerful yet, able to identify thousands of zero-day vulnerabilities over several weeks.  In early April, Anthropic gave 50 initial partners, including the U.S. government, access to Claude Mythos Preview to scan their codebases for vulnerabilities and security flaws. 

The expanded list of organizations with access to Mythos as of today covers power, water, healthcare, communications, and hardware — industries that weren’t “well-represented” in Anthropic’s initial cohort, the company said. Many who will now have access are companies or nonprofits that maintain codebases which other organizations and governments rely upon, Anthropic noted in a blog post.

“What each partner has in common is that a successful attack on their codebase could be catastrophic,” the company said. “For most partners, we estimate that a major attack could affect more than 100 million people, with important ramifications for both global and national security.”

The expanded group includes organizations in countries friendly to the U.S., including Australia, Canada, France, Germany, Italy, Switzerland, the Netherlands, Spain, Belgium, Sweden, India, Japan, New Zealand, and South Korea, according to The Financial Times, citing a person familiar with the matter.

The FT also reported several organizations that have been given access to Mythos, including: U.S.-based identity and security management tool Okta; South Korean companies Samsung, SK Hynix, and SK Telecom; NATO, the U.S.-led military alliance headquartered in Brussels; and the EU’s cyber security agency ENISA. 

TechCrunch has reached out to Anthropic to confirm. 

Anthropic has said it expects other AI companies to soon develop models as capable as Mythos Preview, which is why the firm is racing to establish safeguards within Project Glasswing. 

Since releasing Mythos, rival OpenAI released its own cybersecurity-focused model GPT-5.5-Cyber, which it has rolled out to a large group of partners for testing.

Anthropic, the AI lab behind Claude, has filed confidentially for an initial public offering, the company said in a blog post Monday.

The company, which is valued at close to $1 trillion, submitted a draft registration statement to the U.S. Securities and Exchange Commission for a proposed initial public offering. Anthropic has yet to list the number of shares or set the price. Anthropic said the proposed initial public offering will depend on market conditions and other factors.

The filing comes less than a week after Anthropic raised $65 billion in a Series H funding round that pushed its valuation to $965 billion. The round, which was co-led by Altimeter Capital, Dragoneer, Greenoaks, Sequoia Capital, Capital Group, Coatue, and D1 Capital Partners, attracted a bevy of institutional and strategic investors in anticipation of an IPO.

Anthropic’s confidential filing landed in an already white-hot IPO season that includes SpaceX’s initial public offering that is targeting a $2 trillion valuation. SpaceX is seeking to raise more than $75 billion.

A confidential IPO filing allows a company to begin preparing for a potential public offering without publicly disclosing detailed financial information, risks, or other internal business details. Anthropic will evaluate the IPO privately and without the critical eye of the public. If it follows through, Anthropic will file an S-1 registration document that will contain detailed information about the company’s financials, legal matters, risks, and a breakdown of who holds the most voting power.

Anthropic’s filing also comes as its rival OpenAI continues to raise funding, notably a $122 billion round in March at an $852 billion post-money valuation and prepares for its own IPO. OpenAI is expected to file for an initial public offering, setting the stage for an IPO season that will pit the two largest AI labs against each other and test the market’s resolve and interest in artificial intelligence.

Anthropic, now an AI powerhouse that has landed top-tier enterprise customers, was once considered an underdog in the emerging world of large language models. The startup was founded in 2021 by former OpenAI employees and was seen for years as a distant competitor to OpenAI and its AI chatbot, ChatGPT.

The company has drawn in investors and customers with growing capabilities and a focus on enterprise services. That has translated to eye-popping revenue growth. The company said recently that its revenue run-rate had surpassed $47 billion, up from $9 billion at the end of 2025.

That revenue growth rate could accelerate as Anthropic makes its Mythos model more widely available. Anthropic previewed Mythos in April but has kept access restricted — warning software developers that the model had discovered thousands of high-severity bugs that would need to be fixed before it could be made public.

The generative AI lab is poised to give the European Union’s cybersecurity agency access to Mythos, Bloomberg reported Monday morning, citing anonymous sources.

TL;DR

Salesforce’s stake in Anthropic is now worth approximately $5 billion after investing since early 2023, representing roughly two-thirds of its entire strategic portfolio. The return comes as Anthropic files confidentially for an IPO at a $965 billion valuation.

Salesforce has a stake in Anthropic worth approximately $5 billion after investing repeatedly in the Claude developer since early 2023, Bloomberg reported on Monday. The software company first participated in Anthropic’s fundraising with roughly $50 million and has continued investing in every subsequent round. With Anthropic’s confidential IPO filing also landing on Monday, Salesforce is positioned to realise one of the largest venture-style returns in enterprise software history.

The scale of the return becomes clearer in portfolio context. Salesforce’s total strategic investments across hundreds of companies were valued at $7.8 billion at the end of April, according to SEC filings. That figure predates Anthropic’s latest $65 billion funding round, which valued the company at $965 billion and more than doubled its previous valuation. After the revaluation, the Anthropic stake alone represents roughly two-thirds of Salesforce’s entire strategic investment portfolio.

The early bet

Salesforce Ventures acknowledged in a February blog post that the initial investment was not a consensus trade. “In 2023, betting on a research-heavy startup at a meaningful valuation was not obvious, given there was no market precedent of AI research turning into commercial success,” the firm wrote. Anthropic’s models now power AI features across Salesforce’s product suite, including Slack, and CEO Marc Benioff has said Salesforce will spend $300 million on Anthropic tokens in 2026.

The investment structure gives Salesforce a double exposure. It is both an investor capturing Anthropic’s valuation gains and a customer embedding Anthropic’s models into its own products, including the Agentforce AI agent platform that reached $1.2 billion in annual recurring revenue last quarter. As enterprise AI spending shifts toward agentic platforms, Salesforce benefits regardless of whether its own AI products or Anthropic’s underlying models capture more of the value.

Not the only windfall

Salesforce is not the only company reaping outsized returns from early Anthropic bets. Zoom invested approximately $51 million in Anthropic’s Series C in May 2023 through its Zoom Ventures arm. That stake is now worth approximately $1.3 billion, a roughly 25x return in three years. Accel, which invested in Anthropic’s Series G, has seen its stake more than quadruple in months as the company’s valuation surged past OpenAI’s.

The returns reflect a dynamic that has become defining in the current AI cycle: a small number of model providers, primarily Anthropic and OpenAI, are capturing an extraordinary share of both venture capital and unrealised gains. For Salesforce, the $5 billion Anthropic stake has appreciated more than the combined market capitalisation gains of many of its software acquisitions.

The risks ahead

The $5 billion valuation is on paper until Anthropic’s IPO provides a public market price. Anthropic’s ongoing legal dispute with the Pentagon, which designated the company a supply-chain risk after it refused to grant the military unrestricted model access, represents a material overhang. Anthropic has said the designation could jeopardise billions in revenue.

For Salesforce specifically, the concentration of its strategic portfolio in a single company creates exposure that investors will scrutinise. If Anthropic’s IPO prices below the $965 billion private valuation, or if the stock declines post-listing, the mark-to-market impact on Salesforce’s balance sheet would be significant. Salesforce itself announced the acquisition of Contentful on the same day, continuing a heavy spending pace that has kept its own stock under pressure for months.

But the strategic logic of the original bet has been validated. Salesforce invested in Anthropic before the market had consensus that AI research labs could become commercial businesses. Three years later, Anthropic is filing for a public listing at a valuation approaching $1 trillion, and the $50 million that started the relationship has grown roughly 100-fold.