Bloomberg: "No reliable safe havens." Billionaires have been investing elsewhere. Here's how to get in.

Bloomberg's Marcus Ashworth wrote plainly recently: "No more reliable safe havens."

After all, the S&P fell over 7% from the February peak. Bonds, even with less risk, are barely keeping pace with inflation.

So-called "diversified" portfolios have gotten hit from multiple directions.

Meanwhile, the world’s wealthiest have been setting records in another asset class.

Circumstances are always unique, but after the dot-com bust, it grew roughly 24% annually for a decade. After 2008, roughly 11% annually for 12 years.

Blue-chip art.

Why? It trades globally in multiple currencies, has scarce supply, and has shown near-zero correlation to equities since 1995.*

With Masterworks, 70,000+ investors allocated $1.3B fractionally across 500+ artworks featuring Banksy, Basquiat, and Picasso.

Accredited? You can invest in a diversified portfolio of postwar and contemporary art alongside two other real assets. From 2017-2025, the mix would’ve beat the S&P 500 by 3.1x.

See if you can improve your portfolio performance all in one diversified strategy.

Our subscribers skip the waitlist

^{*According to Masterworks data. Investing involves risk. Past performance is not indicative of future returns. Important Reg A disclosures: masterworks.com/cd}

Today:

• The Vatican and Anthropic Announce Ethics Partnership 

• Huawei Announces New Chip Design Method to Overcome Equipment Bans 

• Microsoft Limits Rival Software Access to Secure Workplace AI Data 

• Google Expands Hidden Labels to Identify AI Images and Video Online 

• Retailer Catalyst Brands to Use Human-Like Robots in Shipping Centers

Faith and Frontier AI – The Vatican Partners with Anthropic on Ethics

Pope Leo XIV Announces Collaboration with Anthropic on AI Ethics 

Pope Leo XIV has officially presented the first encyclical of his papacy, calling for the urgent "disarmament" of artificial intelligence. Highlighting severe concerns over wealth inequality, autonomous weapons, and the concentration of tech power among private corporations, the Pope argued that AI risks weakening authentic human relationships. 

To ensure the technology is guided by strict ethical oversight, the Vatican announced a groundbreaking collaboration with US-based AI startup Anthropic to help navigate the ethical minefields of rapid AI deployment.

The "Magnifica Humanitas" Encyclical (Vatican Official Document) 

The 82-page encyclical, titled Magnifica Humanitas ("Magnificent Humanity"), serves as an authoritative moral teaching on safeguarding the human person in the age of AI. Pope Leo argues that society faces a choice between building a "civilization of love" centered on human dignity or submitting to a "culture of power" driven by unfeeling algorithms. 

Crucially, the encyclical condemns the use of AI in warfare, declaring that the traditional "just war" theory is now outdated because delegating lethal decisions to opaque algorithms removes essential moral responsibility and human control.

Anthropic's Chris Olah on the Pope's Encyclical 

Anthropic co-founder Chris Olah, a self-described atheist, spoke at the Vatican presentation and welcomed the Church's heavy oversight. Olah admitted that AI developers cannot be their own moral compass, as they are inherently driven by internal incentives like pride, ambition, and intense commercial and geopolitical pressures. 

He emphasized that the AI industry desperately needs informed, external critics like philosophers, ethicists, and ancient institutions to answer complex questions about human flourishing and to help keep the industry's worst impulses in check.

Bypassing the Blockade – Huawei's Tau Scaling Law and the Path to 1.4nm

Huawei Unveils Tau Scaling Law to Build 1.4nm Chips Without ASML. Facing strict US sanctions that block access to ASML's advanced EUV lithography machines, Huawei's chip chief He Tingbo unveiled a radical new workaround at the IEEE ISCAS 2026 symposium in Shanghai. Dubbed the "Tau (τ) Scaling Law," Huawei is shifting its focus away from traditional transistor shrinkage (Moore's Law) and is instead minimizing signal transmission time (represented by the Greek letter τ).

To achieve this physically, Huawei introduced "LogicFolding"—a method of stacking 2D circuits into 3D layouts to dramatically shorten the distance data must travel. Huawei claims this multi-level optimization yields a 55% jump in transistor density and a 41% gain in power efficiency. The technology will debut commercially this autumn in the Kirin processor for the Mate 90 series. 

Ultimately, Huawei expects this approach to yield 1.4nm density-equivalent AI and smartphone chips by 2031, establishing a homegrown alternative to Western semiconductor dominance.

The Enterprise Agent War – Microsoft Throttles Databricks in Power BI

Microsoft has quietly escalated its enterprise software war by restricting partners like Databricks from connecting key data-management tools to Power BI. While it may look like a standard corporate integration dispute, the true battleground is over the future of workplace AI agents.

To function effectively, AI agents require a "semantic layer" of trusted, clearly defined business data (e.g., agreed-upon definitions for revenue, churn, inventory, and margins). Microsoft is aggressively pushing its customers to use Fabric and Power BI as their core intelligence stack, undercutting Databricks' ambition to be the primary ecosystem where enterprise data is managed. 

Whoever controls this semantic data layer will ultimately dictate how enterprise AI agents access, learn from, and interpret corporate truths.

🧠RESEARCH

Toward Native Multimodal Modeling: A Roadmap

This paper maps out a guide for building "native" artificial intelligence systems that naturally understand and create text, images, and audio simultaneously. Instead of taping separate programs together, this approach processes all data types in a single, unified system, making future AI much faster and significantly more capable. 

CUA-Gym: Scaling Verifiable Training Environments and Tasks for Computer-Use Agents

Training artificial intelligence to use a computer like a human is difficult because models need realistic practice spaces. This research introduces CUA-Gym, a tool that automatically creates fake websites and tasks for AI to practice on, while reliably scoring their performance so they can learn from their mistakes. 

DVAO: Dynamic Variance-adaptive Advantage Optimization for Multi-reward Reinforcement Learning

When teaching artificial intelligence using multiple different rewards, the training process often becomes unstable and crashes. This paper proposes a new method that dynamically adjusts the training focus in real time. It automatically prioritizes the clearest learning signals while ignoring confusing ones, making the teaching process much more stable. 

📲SOCIAL MEDIA

🗞️MORE NEWS

Google’s AI Media Detection Tools Google expanded its invisible watermarking tools across its search engine and web browser to help people easily spot artificially generated images, videos, and audio. The company also announced that major competitors like OpenAI and Nvidia will start using this same technology to label their own computer-generated content. This team effort aims to make the internet more transparent by creating a shared standard for tracing exactly how media was created.

Figure AI Partners with Catalyst Brands Robotics startup Figure AI signed a deal with retail group Catalyst Brands to deploy human-like robots inside large shipping warehouses. The walking machines will take over physically demanding sorting and packing jobs, freeing the company's human workers to handle more complex tasks. This marks a major step in using physical artificial intelligence to speed up the supply chain for classic retail stores like JCPenney and Brooks Brothers.

Nvidia’s Breakthrough in Image Generation Nvidia researchers created a new software method called "PiD" that drastically speeds up how artificial intelligence generates high-quality pictures. Instead of using a slow, multi-step process to build and then enlarge an image, this new tool translates raw computer data directly into a crisp final picture in one swift move. This breakthrough allows standard home computers to create highly detailed, large-scale images up to six times faster than older systems.

OpenAI Hires New Marketing Chief OpenAI hired Colin Fleming, a former executive at software giant ServiceNow, to take over the marketing efforts for its corporate division. Fleming's main goal will be convincing large businesses to buy and integrate OpenAI's artificial intelligence tools into their daily office work. He replaces the previous marketing chief, who stepped down earlier this year to focus on her health.

SpaceX and Anthropic's $40 Billion Deal Financial documents for SpaceX's upcoming stock market launch revealed a massive $40 billion deal to provide computing power to the artificial intelligence company Anthropic. Anthropic plans to pay SpaceX over $1 billion a month to use its massive data centers, which are required to run incredibly complex software programs. However, the agreement comes with a major catch: either company can cancel the long-term contract at any time with just a 90-day warning.