AI News You Should Know: AGI Benchmarks, Defense AI, and the End of Apps

AI news this week ranged from measuring machine intelligence to military applications to the future of your phone. Here is what matters for business owners.

DeepMind Proposes a New Way to Measure Progress Toward AGI

Google DeepMind released a cognitive framework designed to benchmark how close AI systems are to Artificial General Intelligence. Instead of vague claims, this framework sets specific milestones for what AI should be able to do at each level.

What this means for your business: You do not need to wait for AGI to benefit from AI. The tools available today can already handle scheduling, customer follow-ups, and data entry. Frameworks like this help separate real progress from hype, so you can make better decisions about when to invest.

Berkeley Researchers Crack Open How Large Language Models Think

Researchers at UC Berkeley developed two new algorithms, SPEX and ProxySPEX, that help explain how large language models process information internally. This could lead to more predictable and reliable AI tools.

What this means for your business: More transparent AI means more trustworthy AI. For service businesses using AI chatbots or customer-facing tools, better model interpretability means fewer surprises and more consistent responses to your customers.

The Pentagon Wants AI Companies to Train on Classified Data

The Department of Defense is setting up secure environments where AI companies can train models on classified information. The goal is to build military AI tools that understand sensitive operational data.

What this means for your business: This signals that even the most security-conscious organizations are finding ways to adopt AI safely. If the Pentagon can figure out secure AI training, your business can too. Start with your internal processes and work outward.

Meta Struggles with Rogue AI Agents

A rogue AI agent at Meta accidentally exposed company and user data to engineers who should not have had access. The incident highlights the risks of deploying autonomous AI systems without proper guardrails.

What this means for your business: AI agents are powerful but they need boundaries. If you are automating tasks with AI, make sure you have clear rules about what data the AI can access and what actions it can take. Start small, test thoroughly, and expand carefully.

Tech CEO Predicts Apps Will Disappear, Replaced by AI Agents

Nothing CEO Carl Pei says smartphone apps will eventually be replaced by AI agents that handle tasks autonomously. Instead of opening an app to book a restaurant, you would just tell your AI agent to do it.

What this means for your business: This trend matters for any business with a customer-facing app or online booking system. If customers start interacting through AI agents instead of apps, your business needs to be discoverable and bookable by those agents. Think about how your services appear to AI, not just to humans.

Patreon CEO Challenges AI Companies on Fair Use

Patreon CEO Jack Conte argued that AI companies should pay creators whose content is used to train models. He called the "fair use" defense bogus and pushed for new compensation models.

What this means for your business: If you create content, whether blog posts, videos, or social media, your work may be training AI models right now. This debate will shape future rules around content ownership. For now, keep creating valuable content for your audience and stay aware of how licensing terms evolve.

Mistral Launches Build-Your-Own AI Tools for Enterprises

French AI company Mistral released Mistral Forge, a platform that lets businesses train custom AI models from scratch on their own data. This challenges the one-size-fits-all approach of larger AI providers.

What this means for your business: Custom AI models used to require a massive budget. Tools like Mistral Forge are making tailored AI more accessible. If your business has unique data, like customer service logs or operational records, custom models — combined with an AI Analyst — could give you an edge over competitors using generic tools.

Gemini AI Boosts Productivity in Google Workspace

Google expanded Gemini AI features across Workspace, including email summarization, content drafting, data organization, and meeting tracking. These tools are available to businesses already using Google Workspace.

What this means for your business: If your team uses Google Workspace, you already have access to AI tools that can save hours each week. Start with email summaries and meeting notes. These are low-risk, high-reward ways to get your team comfortable with AI before tackling bigger projects.

The Bottom Line

The tools are here and they are getting better every week. From Google Workspace features you can turn on today to custom AI models you can build on your own data, the barrier to entry keeps dropping. The businesses that start now will compound their advantage over those that wait.

---

Want to see how AI can handle calls, admin, and follow-up for your business? Book a free 15-minute call with our AI team and we will map your biggest opportunities.