Good morning. It's Wednesday, March 4, and we're covering OpenAI’s Pentagon policy reversal, AI’s growing power-grid demands, Google’s Gemini 3.1 Flash-Lite launch, and more.

New here? Keep up with the future of tech, sign up here.

⚡ Gemini 3.1 Flash-Lite Debuts for Scalable AI
Google introduced Gemini 3.1 Flash-Lite, its fastest and most cost-efficient Gemini 3 model for high-volume workloads. Priced at $0.25 per million input tokens and $1.50 per million output tokens, it delivers 2.5x faster time to first token and 45% higher output speed than 2.5 Flash. It scored 1432 on Arena.ai and posted strong multimodal and reasoning benchmarks.

💬 GPT-5.3 Instant Improves Everyday Chat Flow
OpenAI released GPT-5.3 Instant, updating ChatGPT’s most-used model to make conversations more fluid and helpful. The upgrade improves answer accuracy, web search context, and tone while reducing unnecessary refusals and defensive disclaimers. It reduces dead ends and overly cautious phrasing, prioritizing real-world usability over benchmark gains based on user feedback.

🛃 US Weighs 75,000-Chip Cap on NVIDIA Sales To China
US officials are considering limiting NVIDIA to exporting no more than 75,000 H200 AI chips per Chinese customer, potentially tightening the company’s reentry into China. The Trump administration may extend the cap to AMD’s MI325 chips, counting both toward a per-customer limit. The move signals continued US scrutiny of advanced AI exports despite pressure to resume sales.

👨‍⚖️ Supreme Court Lets AI Art Copyright Ban Stand
The US Supreme Court declined to review a ruling that AI-generated art cannot be copyrighted, leaving intact lower court decisions requiring human authorship. The case began with Stephen Thaler’s bid to copyright a 2019 AI-generated image, rejected by the US Copyright Office. Courts upheld the decision in 2023 and 2025, reinforcing that purely AI-created works aren’t copyrightable.

It can - if the AI actually knows your brand, content, and compliance rules. Templafy’s Document Agents combine generative AI with governed workflows so professionals across your org generate accurate, polished, on-brand presentations in minutes.

4M+ users. 30% of document time saved. Works inside PowerPoint, Google Workspace, and Salesforce.

Check out Templafy →

Quick test. One image. Two possibilities. Let’s see if you can still tell what’s real. Is this image real or fake? Answer at the bottom! 👇

MyFitnessPal buys Cal AI, but AI agents may make apps obsolete, showing how vibe coding is reshaping the future of software.

The Recap: OpenAI amended its newly announced agreement with the U.S. Department of Defense to explicitly prohibit the use of its AI systems for domestic surveillance of Americans. The change followed criticism over a March 1 deal that allowed the Pentagon to use OpenAI’s tools for “any lawful purpose,” as rival Anthropic failed to reach similar terms and was labeled a “supply-chain risk” by Defense Secretary Pete Hegseth. OpenAI CEO Sam Altman said the updated language was meant to clarify civil-liberties protections, acknowledging the initial rollout “looked opportunistic and sloppy.”

Highlights:

Forward Future Takeaways:
This episode underscores how AI vendors are becoming direct actors in national-security policy debates, not just technology suppliers. OpenAI’s ability to secure revised language where Anthropic could not suggests contract design — and political timing — may shape which firms gain access to sensitive government work. → Read the full article here. (Paywall)

Ann Davis Vaughan reported that U.S. grid operators in Texas, the mid-Atlantic, and the Midwest have approved roughly $75 billion in transmission upgrades centered on 765-kilovolt (kV) extra-high-voltage lines to serve multi-gigawatt AI data centers. The projects would expand the nation’s 765 kV network from just over 2,000 miles to about 10,000 miles within a few years, as hyperscalers including Meta Platforms, Amazon, and Google scale campuses. A proposed $10 billion “Panhandle Plan” in Texas alone could support up to 24 gigawatts (GW) of data center load, signaling how AI infrastructure is reshaping the U.S. power grid.

AI’s infrastructure race is no longer just about chips and models, it’s redrawing the physical map of the U.S. power grid. The scale of 765 kV buildouts suggests regulators and utilities now see hyperscale data centers as anchor tenants capable of underwriting multibillion-dollar transmission bets. → Read the full article here. (Paywall)

Aida Baradari announced Spectre I, a pocket-sized device from startup Deveillance that claims to prevent nearby microphones from recording speech. In posts on X, Baradari said the device scans for microphones and emits AI-generated “cancellation signals” designed to render speech unintelligible to phones, laptops, and smart speakers.

Unlike conventional audio jammers that rely on high power, Spectre I is described as targeted, portable, and based on “novel physics research.” The company has opened pre-orders and is recruiting hires, though no independent technical validation was shared in the announcement. → Read the full article here.

Qualcomm Bets Big on Robotics: CEO Cristiano Amon says robots will become a larger opportunity within two years, following January’s launch of its Dragonwing robotics chip.

43% Want Career Changes: A FlexJobs survey of 4,000 U.S. workers finds 43% aim to switch fields in 2026, driven by AI fears and layoffs—yet quits remain low at 2%.

Apple Unveils M5 Pro, Max: New 3nm chips fuse two dies, add an 18-core CPU and up to 40-core GPU, boosting AI and graphics up to 4x. Preorders start March 4; sales begin March 11.

Claude Code Adds Voice Mode: Anthropic rolls out push-to-talk voice input to ~5% of users, expanding in weeks. Transcripts stream into prompts, with no extra cost on paid plans.

ChatGPT Uninstalls Spike 295%: After OpenAI’s DoD deal, U.S. app deletions surged 295% on Feb. 28, while Claude downloads jumped up to 88% and hit No. 1 on the App Store.

The image is real. Tune in next week for another Real or AI test.

Postiz: Schedule posts, use AI tools, track analytics, and grow your audience.
PicWish: Free AI photo editor for background removal, enhancement, and portraits.
Questflow: Orchestrate AI agents to automate tasks, insights, and rewards.

❤️ Love Forward Future? Share your referral link with friends and colleagues to unlock exclusive perks! 👉 Get your link here.
📢 Want to advertise with us? Reach 1M+ AI enthusiasts, tech leaders, and decision-makers. Just reply to this email.

Thanks for reading today’s newsletter. See you next time!
Matthew Berman & The Forward Future Team

🧑‍🚀 🧑‍🚀 🧑‍🚀 🧑‍🚀 🧑‍🚀