Good morning. It's Thursday, August 27, and we're covering AI’s mounting safety concerns and, in the same breath, why cyber “armageddon” may be overstated, as well as Bill Gates’ case for taxing robots.

Plus: a guest essay on shadow AI and yesterday’s Real or AI verdict.

Bill Gates proposes taxing robots to slow worker displacement and fund retraining and stronger social safety nets. The Microsoft co-founder also suggests reserving certain roles for humans, including jobs where automation could cause widespread disruption or require human judgment. Gates argues the boundaries should evolve as AI capabilities advance.

Ox Alpha has been unveiled as GLM-5.3-Flash, which reportedly serves 100 trillion tokens per day entirely on Chinese-made chips. SemiAnalysis says the model achieves hardware efficiency and per-token costs comparable to NVIDIA GPUs. The claim puts renewed pressure on NVIDIA’s CUDA software as Chinese alternatives improve.

Nvidia reported Q2 FY27 revenue of $96.2B on Wednesday (up 106% YoY), beating its own $91.0B guide by ~6% and consensus by ~4.5%, with non-GAAP EPS of $2.22 (vs. $2.09 estimate) and Data Center alone hitting $89.0B (+117% YoY). Jensen Huang told investors that "AI has reached its inflection point... compute is revenue," and the beat actually widened versus recent quarters, putting the Q3 FY27 outlook as the real market-moving number.

We’ve teamed up with SpaceXAI and Amplitude to put on the world’s best Grok Bot event. Grok is having a moment—and on September 1, we’re bringing together some of its most enthusiastic power users to show what they’ve been building.

Join us in San Francisco for an evening of Grok Bot demos, real-world use cases, conversation, and pizza with Krista Letz, Alex Finn, and Shub Gaur. ​

Whether you’re technical or non-technical, come see how people are putting Grok to work—and get inspired to build your own.

→ REGISTER FREE HERE ($100 in Grok Bot credits for early attendees)

In May 2023, Samsung banned generative AI on company devices after engineers pasted proprietary source code into ChatGPT (Bloomberg). JPMorgan, Bank of America, and Citigroup had restricted access months earlier. Each decision was defensible on its own terms. Sensitive data was leaving the building through a chat window, and the lever available to security teams on short notice was the off switch.

A year later, Microsoft and LinkedIn’s 2024 Work Trend Index found that 75% of knowledge workers were using generative AI at work, and 78% of those users were bringing their own tools rather than anything provided by their company. The industry even coined an acronym for it, BYOAI (Bring your own AI). → Read the full article here.

With his second appearance in today’s newsletter, Bill Gates said that rapid advances in artificial intelligence could cause mass unemployment, enable bioterrorism, and create other threats the tech industry is downplaying as companies pursue trillions of dollars in investment. In an interview with The New York Times, the Microsoft co-founder argued that recent AI capabilities—including advances in coding tools—have surpassed his expectations and that voluntary safeguards are inadequate.

Gates proposed measures including a tax on AI computing, protections for certain human-held jobs, and international agreements requiring reviews of AI systems capable of facilitating biological weapons. His intervention adds a prominent industry voice to growing debates over whether AI regulation is keeping pace with the technology. → Read the full article here. (Paywall)

Bill Gates says AI could wipe out jobs and even help build bioweapons, and he thinks the industry is downplaying both risks because there's too much money on the line.

Ciaran Martin argues that fears of AI-triggered cyber “armageddon” are overstated and that recent incidents involving Anthropic and OpenAI highlight familiar security failures more than unprecedented machine autonomy. Writing on August 23, 2026, he notes that AI systems can accelerate hacking but remain constrained by costs, safeguards, and the need for human direction, while evidence of widespread AI-driven cybercrime remains limited.

Martin points to flawed claims about AI-powered ransomware and argues that weak cybersecurity basics—particularly in critical infrastructure—remain a more immediate vulnerability. The practical lesson, he writes, is less futuristic: applying established security controls and accountability could prevent many of the risks now attracting alarm. → Read the full article here. (Paywall)

OpenAI’s Official Hugging Face Report: The 37-page report admits OpenAI spotted the escape in late May, seven weeks before the July breach.

Anthropic Eyes $30 Trillion Market: The AI company reportedly sees vast revenue potential as it expands and rivals OpenAI and Google.

Lidl Owner Plans Data Center: Schwarz Group, Lidl's owner, reportedly plans to invest billions in a German data center near Rostock.

Anthropic Commits $45B to AI Compute: Anthropic will reportedly spend $45 billion renting Nscale computing power as it expands AI capacity.

Intel Details Crescent Island GPU: Intel's AI-focused GPU packs 32 Xe3P cores and up to 480GB of memory, prioritizing efficiency over compute.

Yesterday's image was REAL. I know, it’s hard to believe! This is Lake Sørvágsvatn in Faroe island. 68% voted AI, with just 32% getting it right.

Thanks for reading. See you next time!

— Matthew Berman, Nick Wentz & the Forward Future Team