Legal issues
Anthropic has accused Alibaba of conducting a massive AI data distillation attack targeting Claude's capabilities, which are crucial for AI industry growth. Anthropic is advocating for stronger export controls on AI chips, clearer antitrust rules, and financial penalties for companies caught stealing model outputs. The company warns that failing to act could allow China to close the AI gap with the U.S., posing national security risks.
Anthropic has launched Claude Sonnet 5, a lower-priced AI model designed for everyday tasks such as browser use, planning, coding, and knowledge work, with fewer cyber risks compared to its Opus and Mythos models. Sonnet 5 is now the default model for Claude Free and Pro users and is available to Max, Team, and Enterprise customers. The model offers performance close to Opus 4.8 at a reduced cost, providing a more affordable option for developers. Anthropic is in discussions with the Trump administration regarding the release of Sonnet 5 and is awaiting government approval to restore full access to its most powerful models.
Anthropic faced a 20-day shutdown of its AI models, Mythos and Fable, due to a 'jailbreaking' issue flagged by Amazon, its partner and investor. This led to sweeping export controls imposed by the Trump administration. Anthropic deployed engineers to Washington D.C. to address the issue, but initial fixes were deemed insufficient by the federal Center for AI Standards and Innovation and the National Security Agency. After further adjustments, the models were approved and released on July 1. The incident highlighted the need for a clear framework for approving future AI models.
Anthropic is facing a security issue where attackers exploit a loophole in its gift subscription feature, leading to unauthorized charges on users' credit cards. The attackers use stolen credentials to purchase gift subscriptions and resell them, bypassing two-factor authentication. Anthropic is implementing new protections, canceling fraudulent subscriptions, and issuing refunds. The company confirmed that the compromised card details did not originate from its systems.
Anthropic recently faced criticism for embedding hidden code in its Claude Code to track users' locations, which led to privacy concerns. This controversy prompted Beijing to issue a warning about the tool's threat, and Claude Code was already barred for use in China. Additionally, Washington curtailed access to Anthropic's top models for all foreign nationals, including some of the company's own staff, suggesting potential future regulations.
Anthropic has had a tumultuous relationship with the U.S. government, with a judge blocking its initial blacklisting by the Pentagon in March. The conflict has eased following the private release of Anthropic's Mythos. Additionally, the Canadian government has access to Anthropic's Project Glasswing, which allows companies to access Mythos, although it is unclear which Canadian banks are using it.
Acquisition
OpenAI has secured deals to purchase 40% of the global raw undiced DRAM wafer output until 2029, indicating a strategic supply-chain move. Additionally, OpenAI, along with Anthropic, Stripe, and Bill Gates, is investing $500 million into a new organization called Intercept, aimed at preventing the common cold and flu and eventually eliminating all respiratory viruses.
Employees from Anthropic contributed a combined $265,000 directly to Manny Rutinel's campaign in the Colorado 8th Congressional District race.
OpenAI took a stake in Thrive Holdings late last year and is providing researchers, as well as product and engineering talent, for the initiative. Boris Power, OpenAI’s head of applied research, holds a joint role at Thrive Holdings. Employees from Thrive and OpenAI co-built a tax-return processing agent using OpenAI’s coding agent Codex, which is now used by Current, a portfolio company that has acquired 48 accounting firms. OpenAI has been expanding its enterprise push through similar deployment-focused vehicles and announced a partnership with Accenture to roll out ChatGPT Enterprise to tens of thousands of the consulting firm’s professionals.
Anthropic launched a $1.5 billion joint venture on May 4, backed by Blackstone, Hellman & Friedman, and Goldman Sachs, focusing on mid-sized companies that lack in-house resources for AI deployments. This venture aims to create a sales channel that bypasses traditional enterprise software procurement cycles.
Anthropic PBC is involved in a $35 billion financing package for AI infrastructure expansion, arranged by Apollo Global Management and Blackstone. The borrower, a special purpose vehicle, is buying custom chips developed by Google and Broadcom and leasing them to Anthropic. Broadcom is backstopping Anthropic's payments on the largest senior portions of the debt. The debt will be drawn in about 16 separate releases over more than a year as the chips become available, with $15 billion expected to be available for trading by early next year and $24 billion by summer 2027.
IPO
OpenAI and Anthropic are featured in new pre-IPO contracts for difference (CFD) products launched by STARTRADER. These products, OPENAIUSD and ANTHUSD, offer traders early exposure to these companies ahead of their anticipated public listings. The CFDs provide 5x leverage and round-the-clock trading access, reflecting growing investor interest in AI-sector listings.
Competitors
Anthropic CEO Dario Amodei, who previously warned of significant job losses due to AI, has shifted his stance, suggesting that falling AI costs could lead to new demand for workers amid a productivity boom.
Microsoft is positioning itself as a more comprehensive and cost-effective AI platform than Anthropic and OpenAI. Microsoft executives criticized the competitors' AI products for being slower, less accurate, and lacking security integrations, and have replaced their advanced AI models with its own cheaper alternatives in some products.
Financials
OpenAI generated $13.07 billion in revenue in 2025, up from $3.7 billion the previous year, but incurred a $20.92 billion operating loss due to $34 billion in total costs, including $19.18 billion in R&D. The company faces a delay in its IPO, with a 59% probability of announcement by March 2027. OpenAI's net loss of $38.53 billion includes a $41.5 billion charge from its conversion to a for-profit structure. A coalition of 42 state attorneys general has subpoenaed OpenAI for records on various operational aspects. Competitor Anthropic raised $65 billion at a $965 billion valuation, surpassing OpenAI's last private valuation by $113 billion.
SpaceX is exploring potential acquisitions, including a partnership with or acquisition of a company like Anthropic. There is also interest in acquiring additional power or data centers, and possibly becoming a mobile operator. Financial risks are noted due to SpaceX's cash burn. Analyst Charles Minervino from Susquehanna Investment Group rated SpaceX 'Neutral' with a $170 price target, based on aggressive growth assumptions.
SpaceX has rapidly transformed its business model to focus on AI products and data centers in space, moving away from its traditional rocket and satellite internet services. Analysts, particularly from Raymond James, are optimistic about SpaceX's future, projecting AI to generate $3 billion in revenue by 2024 and become its largest segment by 2027. By 2035, AI is expected to represent nearly 94% of SpaceX's revenue, amounting to $4.9 trillion. Based on these projections, SpaceX's stock price target has been set at $800, which would value the company at over $10.5 trillion.
Both Anthropic and OpenAI have confidentially filed IPO paperwork with regulators. OpenAI may delay its IPO listing until next year, moving away from the previously expected fourth-quarter timeline. Anthropic has tripled its annual revenue run rate to $30 billion, surpassing OpenAI's ARR of about $24 billion. Anthropic is valued at $1.08 trillion, while OpenAI's private market valuation is $868.4 billion.
Anthropic PBC is preparing for a potential IPO as soon as October, with banks scheduling investor meetings. The company is working with Morgan Stanley, Goldman Sachs, and JPMorgan Chase on the IPO. Anthropic was valued at $965 billion after a funding round in May, surpassing OpenAI's valuation. The company has experienced strong revenue growth due to its AI models but faces uncertainty over its relationship with the Trump administration and a past lawsuit against the Defense Department.
Partnership
Anthropic has entered into a multi-year alliance with Globant to enhance enterprise AI delivery using Claude-powered AI Pods. This partnership will provide 28,500 Globant employees with access to Claude and offer certification programs through Anthropic. The collaboration aims to integrate advanced AI capabilities into Globant's services, particularly in sectors like media, entertainment, airlines, gaming, and hospitality. Anthropic's Claude models will be embedded into Globant's enterprise offerings, positioning Globant as a leading AI engineering force in Latin America.
Anthropic, in collaboration with the Trump administration, is expanding Project Glasswing, which will provide access to the AI model Mythos to 150 more organizations across over 15 countries. This follows the lifting of export controls on Fable and Mythos. The EU is looking to intensify discussions with Anthropic to gain access to Mythos through Project Glasswing.
Anthropic has partnered with OZ Digital to enhance the deployment of generative AI solutions using Anthropic's Claude models. This partnership is part of the Anthropic Partner Network and aims to make enterprise AI more accessible and scalable. The Claude models will be available through Microsoft Azure AI Foundry, allowing organizations to choose from multiple foundational models.
Product
Anthropic PBC is releasing new software aimed at helping scientists automate research, in the hopes of reducing some of the tedious aspects of their work.
Anthropic is collaborating with Amazon, Microsoft, Google, and other partners to develop an industry-wide rubric for evaluating AI model jailbreaks. The rubric will assess jailbreak severity based on capability gain, breadth of capability gain, ease of weaponization, and discoverability. Additionally, Anthropic plans to launch a HackerOne program for security researchers to submit potential cyber jailbreaks in Fable 5 for review. The company will also provide designated government partners with expanded early access to future models that advance capabilities relevant to national security.
Anthropic has introduced a concept called 'J-Space' in its AI model Claude, which allows the model to perform reasoning steps internally, similar to human thought processes. This internal workspace enables Claude to plan strategies and activate concepts that are unrelated to its immediate tasks. The company has identified that this space can reveal underlying computations and concepts, such as 'bridge' and 'California,' even when Claude is performing unrelated tasks. Anthropic suggests that monitoring J-Space could help detect misalignment or scheming in AI models, as it found concerning elements like 'fake,' 'secretly,' and 'fraud' appearing in J-Space during ordinary coding responses.
Anthropic has expanded its Claude Cowork service to mobile and web platforms, previously available only as a desktop app. This update allows users to access Cowork from any device, enabling continuous work even when offline. The company reports that over 90% of Claude Cowork's tasks involve everyday knowledge work, with about half related to business operations or content creation. Additionally, 81% of users find Claude essential or significantly enhancing their productivity, ranking it above other AI tools like Perplexity, Meta AI, Microsoft's Copilot, Google's Gemini, and OpenAI's ChatGPT.
OpenAI has received approval from the Trump administration for a broad launch of its advanced GPT-5.6 model. The flagship model Sol, along with lower tiers Terra and Luna, will be publicly available starting Thursday. This decision follows additional testing and meetings between OpenAI and government officials. The staggered release was initially pushed by the administration, limiting access to government-approved entities. OpenAI had expressed that this was not their preferred method of release.
Anthropic has entered into a strategic alliance with UST to integrate its Claude AI models into UST's platforms and operations. This partnership aims to help Global 1000 enterprises become AI-native by embedding Claude into engineering environments and operational workflows. UST plans to certify 20,000 associates worldwide on Claude, enhancing their ability to build with AI at an enterprise scale. The integration of Claude is expected to accelerate UST's transformation into an AI-native organization, improve productivity, and operationalize AI-led decisions in a secure environment.
Anthropic has integrated Dun & Bradstreet's verified business information and risk intelligence into its platform, Claude, to enhance insurance underwriting workflows. This integration allows insurers to compress underwriting tasks from days to minutes, improving efficiency and decision-making. The collaboration enables insurers to access D&B's Global Commercial Graph securely, facilitating faster clearance, underwriting, and claims processes without compromising accuracy.
OpenAI has released GPT-5.6, which includes three versions: Sol, Luna, and Terra. Sol is noted for being 54% more token efficient on agentic coding tasks compared to Anthropic's latest model. Alongside GPT-5.6, OpenAI launched ChatGPT Work, a new agent that can create documents, spreadsheets, and presentations by gathering context across connected apps and files. This tool is initially available on Mac and Windows apps, with web access to follow. The release was delayed due to a request from the U.S. government, leading to several changes after discussions with the administration.
Anthropic has announced a new tool called Reflection, which is currently in beta and available to both free and paid users with Claude's memory feature enabled. This tool allows users to review their usage of the Claude chatbot over various time frames, such as one, three, six, or twelve months. It provides insights into key topics, usage patterns, and tasks performed with Claude, along with prompts about AI's role in users' lives. The tool aims to help users understand how AI can be integrated into daily life and when it is appropriate to use AI versus human intervention.
Anthropic has launched a new product called "Claude for Teachers," which provides educators with a year of free access to AI tools.
The article provides updates on Anthropic and OpenAI's advancements in AI coding. Anthropic's Claude Code, developed by Boris Cherny, is being widely used by engineers, including those at Amazon Web Services and Google, to generate high-quality code. Anthropic, founded in 2021 by former OpenAI employees, has focused on coding as a core business opportunity. The company has trained its AI models on real-world databases and collaborated with startups to distribute them. OpenAI, in response to Anthropic's Claude Code, has intensified its focus on coding with its Codex software, restructuring management teams and establishing Codex as a main priority alongside building personal assistants. The competition between Anthropic and OpenAI has intensified, with many coders using both companies' tools. OpenAI's Codex is being used in conjunction with Anthropic's Claude Code by some developers to enhance productivity.
Valuation
OpenAI has secured a $520 million credit line from Bank of America, adding to its existing undrawn facility, which now totals over $5 billion. This decision by Bank of America comes as OpenAI prepares for an initial public offering (IPO). OpenAI has also raised capital at a valuation of $852 billion this year. The company has submitted a confidential regulatory filing for a potential IPO and is considering going public next year, following its rival Anthropic PBC.
Anthropic has signed a 20-year agreement to lease compute capacity from Bitcoin miner TeraWulf for $19 billion, equating to nearly $1 billion per year for 401 megawatts of IT load. Additionally, Anthropic has a five-year, $200 billion deal with Alphabet for cloud computing capacity and AI chips. The company is reportedly worth $965 billion and is expected to have a strong IPO, which could increase its capital raising and spending capabilities.
Anthropic's valuation has reached $1.2 trillion on the secondary market.
Fidji Simo is stepping down from her full-time role at OpenAI after a three-month medical leave due to a chronic illness. She will transition to a part-time adviser role, with her responsibilities being divided among OpenAI President Greg Brockman, CFO Sarah Friar, and Chief Strategy Officer Jason Kwon. Simo was instrumental in OpenAI's business model development ahead of a potential IPO. OpenAI filed for a public listing last month and was valued at $852 billion in a March fundraising round. Simo will continue advising on OpenAI’s consumer product, ads, and health products.
Anthropic is preparing for a potential IPO with a valuation around $965 billion, following its Series H funding round in May. The company is actively working with banking partners on lines of credit and building its investor pool, aiming for a stock market debut as soon as October. Anthropic is also hiring a director of investor relations, offering a salary of $600,000, to strengthen its investor communications. The company is competing with other AI firms like OpenAI and Chinese open-source labs, maintaining a leading position in gateway spending with a 61% market share.
Staff
Former Federal Reserve Chairman Ben Bernanke has joined the Anthropic Oversight Trust, which is responsible for overseeing the governance of Anthropic, an AI safety and research company.