5 May 2026 PM: OpenAI and Anthropic Both Bet on Finance AI as Chrome’s Hidden Model Raises Alarms

Two of the most powerful AI labs converged on the same target in a single day: the finance department. OpenAI and PwC are teaming up to reimagine the CFO’s office, while Anthropic is pitching agents directly to insurers and banks. And if you use Chrome, there is a good chance a 4GB AI model has already arrived on your machine without asking.

OpenAI has announced a collaboration with Big Four firm PwC to bring AI agents to enterprise finance workflows. The partnership covers forecasting, financial controls, budget modelling, and audit preparation. The goal is to help CFOs and their teams automate the kind of repetitive, high-stakes work that currently consumes significant hours across any finance function.

The timing matters for UK businesses. PwC has one of the largest advisory networks in Britain, and any AI tools it embeds in its services will reach thousands of mid-sized companies, not just multinationals. Finance teams working with accounting firms should expect to encounter these tools sooner than most AI adoption timelines suggest.

The collaboration does not yet specify a packaged consumer product, but OpenAI’s enterprise API is available to any developer today. SMBs looking to experiment with AI-assisted finance do not need to wait for PwC. The capability is accessible now through OpenAI’s platform, and the PwC partnership is likely to accelerate the appetite for it across the UK accountancy sector.

Anthropic has moved in the same direction, publishing a dedicated post on how its Claude models can serve as agents in financial services and insurance. Use cases highlighted include document analysis, regulatory compliance support, customer query handling, and underwriting assistance. The announcement came within hours of the OpenAI-PwC news, and the overlap is not coincidental.

Anthropic’s positioning emphasises safety and auditability, qualities that matter significantly to regulated industries. Banks and insurers operating under FCA oversight need AI systems whose reasoning can be logged, reviewed, and challenged. Anthropic is marketing those features explicitly, which suggests it has been talking directly to compliance teams at UK financial institutions.

The fact that both labs are targeting finance on the same day signals where the next wave of enterprise AI deployment is heading. Finance has the budgets, the data volumes, and the regulatory pressure to justify AI at scale. Expect Google, Microsoft, and the major ERP vendors to announce competing moves in the weeks ahead.

Google has been quietly installing a 4GB AI model on Chrome users’ devices without prompting or notification. A privacy researcher discovered the model, referred to internally as Nano, stored inside Chrome’s application data folder. The download happens silently via Chrome’s auto-update mechanism, and there is no visible opt-out prompt in the browser interface.

For most users the first sign of anything unusual is unexpected storage consumption. The model is used to power on-device AI features in Chrome, including AI-assisted text suggestions and summarisation. Google’s argument is that keeping the model local improves privacy by avoiding server-side processing. Critics point out that installing a 4GB file without consent is itself a privacy issue, regardless of what the file does.

UK users have specific protections under GDPR and the Data Protection Act 2018. Whether Chrome’s behaviour meets the standard of transparency required by those laws has not been tested yet, but privacy campaigners are calling for an ICO investigation. For now, Chrome users can check their device by searching for “Optimization Guide On Device Model” in their file manager or storage settings.

Voice AI company ElevenLabs has disclosed $500 million in annual recurring revenue and revealed a new set of high-profile investors including asset manager BlackRock, actor Jamie Foxx, and sports executive Eva Longoria. The ARR figure makes ElevenLabs one of the fastest-growing enterprise software companies currently operating.

The milestone matters because it confirms that voice AI has moved well past the experimental phase. ElevenLabs is now embedded in podcast production pipelines, e-learning platforms, gaming studios, and enterprise customer service operations. UK businesses using voice interfaces are increasingly likely to be running on ElevenLabs infrastructure, often without realising it.

The involvement of BlackRock as an investor is a useful signal. Institutional capital at that scale tends to reflect a long-term strategic position rather than speculative backing. Voice AI, in BlackRock’s view, is now an investable asset class.

Stuart Russell, the Berkeley professor whose AI textbook is standard reading at most UK universities, has testified in the Musk-versus-OpenAI case that governments need to intervene to prevent an uncontrollable AGI arms race. Russell is serving as the sole expert witness for Elon Musk in the ongoing legal dispute over OpenAI’s for-profit transition.

His argument is that competitive pressure between frontier labs creates incentives to accelerate development faster than safety research can keep pace. The case itself turns on governance and corporate structure, but Russell’s testimony puts the broader AI safety debate on the record in a US federal court for the first time.

UK policymakers following the case should note that a credible, mainstream AI safety researcher, not a fringe voice, is now arguing publicly for external restraint on frontier labs. That framing is likely to carry weight in parliamentary discussions about the AI Opportunities Action Plan and the government’s approach to frontier model regulation.

Krutrim, India’s first AI-focused unicorn, has announced a pivot away from building proprietary language models and toward cloud services, following a period of layoffs and limited product progress. The shift is a significant recalibration for a company that raised its valuation on ambitious model-building plans.

Building and maintaining frontier AI models requires infrastructure investment that only a small number of companies globally can sustain indefinitely. Krutrim’s pivot is likely to be the first of several similar adjustments from well-funded startups that raised at high valuations on model-building ambitions. The competitive advantage in AI increasingly lies in proprietary data and workflow integration, not in the underlying models. UK founders and investors in the AI space should factor this into their planning.

Here is everything else worth knowing from today’s AI news.

• Image AI models now drive app growth at 6.5x chatbot rates — Appfigures data shows visual model launches generate far more downloads, but most fail to convert that spike into sustained revenue. TechCrunch
• Jensen Huang says AI is creating an enormous number of jobs — Nvidia’s CEO pushed back on automation fears at a public event, arguing AI tools are net job creators rather than replacements. TechCrunch
• Cerebras on track for a $26.6 billion IPO — The AI chip maker, a close OpenAI partner, is advancing toward a blockbuster public listing that would be one of the largest tech IPOs of the year. TechCrunch
• Sierra raises $950 million for enterprise AI customer service — The round gives Sierra more than $1 billion to pursue its goal of becoming the global standard for AI-powered customer interactions. TechCrunch
• Meta to use AI to identify underage users from height and bone structure — A visual analysis system is being tested in select countries, with Meta saying it will move toward a broader rollout. TechCrunch
• CopilotKit raises $27 million for app-native AI agents — Series A led by Glilot Capital, NFX, and SignalFire backs tooling for developers embedding AI copilots in their own applications. TechCrunch
• OpenAI publishes technical breakdown of its low-latency voice AI infrastructure — A detailed post covers how OpenAI routes and processes voice at scale, useful reading for developers building voice applications. OpenAI
• The AI Product Graveyard is growing — A community-maintained list of discontinued AI tools is gaining attention on Hacker News, a reminder that the current boom is producing casualties alongside winners. Tool Directory

This is a daily news update for informational purposes only. AI products and policies change rapidly. Verify details directly with providers before making decisions. Nothing here is financial or legal advice.

AI Daily is Cristoniq’s afternoon update on developments in artificial intelligence, published every weekday afternoon.