This week in AI: Anthropic surpasses OpenAI in valuation after a $65 billion Series H, Claude Opus 4.8 ships with better error detection and a cheaper fast mode, Pope Leo XIV’s first encyclical raises questions about AI transparency in institutional writing, Microsoft redesigns Copilot for Office, and Forbes China spotlights 50 AI companies reshaping global productivity.
1. Claude Opus 4.8: Same Price, Fast Mode 3x Cheaper
Anthropic launched Claude Opus 4.8 on May 28, featuring a fast mode that is now three times cheaper, 2.5x faster than standard mode, and an error-detection capability that is approximately four times more likely to flag code flaws before they pass unremarked. The model adds effort control (users select how deeply Claude reasons), dynamic parallel workflows in Claude Code, and a Messages API update that supports mid-task instruction changes without breaking the prompt cache.
Why it matters: When an AI writes code or builds reports without supervision, the risk is not failure, it is silent failure. A model that raises its hand and says “this might be wrong” reduces that risk, and a cheaper fast mode lowers the cost of deploying these agents at scale.
2. Anthropic Closes $65B and Surpasses OpenAI in Valuation
On May 28, Anthropic announced a Series H of $65 billion, valuing the company at $965 billion post-money, led by Altimeter, Dragoneer, Greenoaks, and Sequoia. That puts Anthropic above OpenAI’s approximately $852 billion valuation, making it the most valuable private AI company in the world. Micron, Samsung, and SK hynix joined as strategic infrastructure partners.
Why it matters: When the provider behind your tools is worth nearly $1 trillion, it is clear that enterprise AI is not a passing trend. The addition of memory and chip manufacturers confirms that the next competitive battle is not about models; it is about physical compute capacity.
3. The Pope’s First AI Encyclical
Pope Leo XIV published “Magnifica Humanitas” on May 25, a 42,300-word document warning about AI risks and calling for stronger regulation. Researchers using the Pangram detector flagged approximately 46% of the content as potentially machine-generated, with some opening chapters reaching 62%, and identified stylistic markers linked to Anthropic’s Claude model. The Pope presented the document alongside Christopher Olah, a co-founder of Anthropic.
Why it matters: This is a case study in AI detection and transparency that any marketing or content team should track when publishing “human” material. If the most traditional institution on the planet uses AI in drafting without declaring it, the pressure for disclosure standards will arrive at companies, media, and governments within months.
For sales teams tired of cold leads, slow customer responses, and manual processes, Dapta is the ultimate tool.
Dapta is the leading platform for creating AI sales agents specifically designed to increase inbound lead conversion. Respond to your leads in less than a minute with voice AI and WhatsApp that converts.
If you want your team to sell more while AI handles the complex stuff, you have to try it.
4. Microsoft Redesigns Copilot
Microsoft began rolling out a redesigned Copilot in Word, Excel, PowerPoint, and Outlook in late May, with a cleaner interface and a single, task-aware entry point. It had to reverse a floating button it had forced on users after a wave of complaints that it covered data and could not be hidden. The company admitted the forced rollout was a mistake, though engagement increased regardless.
Why it matters: Microsoft wants AI to feel like a native Office control rather than a chatbot stuck on top, because if your team lives in Office, that reduces daily friction. The uncomfortable figure underneath: only about 3.3% of Microsoft 365 users pay for Copilot, so adoption remains well below what was promised.
5. Forbes China Reveals Its AI TOP 50
On May 17, Forbes China published its “2026 AI Tech Enterprises TOP 50” selection, led by DeepSeek, Kimi (Moonshot), Qwen (Alibaba), and StepFun. DeepSeek V4 (May 2026) migrates inference completely from CUDA to domestic compute, pushes context windows past one million tokens, and cuts long-context costs by more than 50%. Kimi K2.6 scores 80.2% on SWE-Bench, positioning itself as an enterprise productivity tool.
Why it matters: China is producing tools that are competitive on cost and performance, well-suited for companies seeking cheaper alternatives or better data sovereignty. Many of these tools are already used globally in e-commerce, operations, and customer service, eroding the cost advantage the West took for granted.
Bonus: Content of the Week
The era of thousands of small bootstrapped companies is coming. One favorite is Chatbase, which reached $10 million in annual recurring revenue without outside capital.
