Google's Gemini Robotics-ER 1.6, Anthropic vs OpenAI investors, and the deepfake nudes crisis hitting 90 schools

600 students. 90 schools. And the problem is getting worse, not better. That's the headline from a new Wired investigation into AI-generated deepfake nudes in schools - and it's a number that should stop you in your tracks. But that's not the only story worth your attention today. Google just upgraded the brain it's putting inside robots, and the Anthropic vs OpenAI investor drama is heating up in ways that reveal a lot about where the AI industry is actually heading. Let's get into it.

Google DeepMind's Gemini Robotics-ER 1.6 is described as the 'cognitive brain' of robots. But what does the 'ER' in Gemini Robotics-ER actually stand for?

The answer is hiding near the bottom of today's newsletter... keep scrolling. 👇

Google DeepMind yesterday released Gemini Robotics-ER 1.6, a significant upgrade to the model that acts as the high-level reasoning layer for robots operating in physical environments. Think of it as the part of the robot that figures out what to do - not the part that actually moves the arm.

The model specializes in four areas that matter enormously for real-world robotics: visual understanding, spatial reasoning, task planning, and success detection. That last one is interesting - the ability to recognize whether a task actually worked is something humans do naturally but has been surprisingly hard to encode in AI systems.

One standout capability in the 1.6 release is instrument reading - the ability to visually interpret gauges, dials, and displays in industrial or lab environments. That's the kind of fine-grained visual understanding that opens doors to real deployment in manufacturing and scientific settings. For anyone building physical AI products right now, this is worth a close look.

Read the full story →

A joint investigation by Wired and Indicator has found that nearly 90 schools across the world have been impacted by AI-generated deepfake nude images, with around 600 students affected. And critically - the report finds no signs of the problem going away.

So-called 'nudify' apps - tools that strip clothing from photos using AI - have proliferated online with very little enforcement. Most are free or cheap, require no verification, and are trivially easy to find. The result is that students, predominantly girls, are being victimized by classmates using tools that cost nothing and take seconds to operate.

We've covered child safety and deepfakes before, but this investigation puts hard numbers on a crisis that's often discussed in the abstract. Schools are largely unprepared, legal frameworks are patchy, and the platforms hosting these tools have been slow to act. This is a digital safety failure at every level - tech companies, platforms, and policymakers alike.

Read the full story →

Here's a number that tells you a lot about where the AI industry stands right now: to justify OpenAI's most recent funding round, investors reportedly had to assume an IPO valuation of $1.2 trillion or more. Meanwhile, Anthropic is sitting at a $380 billion valuation - and some backers who invested in both companies are starting to think that looks like the better deal.

TechCrunch reports that the gap between the two valuations - and the assumptions required to justify them - is prompting real second-guessing among investors who have backed both companies. The concern isn't that OpenAI is struggling. It's that Anthropic's trajectory, particularly after the Mythos model release and its deepening enterprise relationships, makes OpenAI's price tag look increasingly hard to defend.

This is worth watching closely. Investor sentiment at this level tends to shape what gets funded, what gets built, and who ends up setting the agenda for the whole industry. If Anthropic continues to close the gap on model quality while OpenAI carries a much heavier valuation burden, the competitive dynamics could shift faster than most people expect.

Read the full story →

The NAACP filed a federal lawsuit yesterday accusing Elon Musk's xAI of illegally spewing toxic pollutants from its data center into Black neighborhoods on the Tennessee-Mississippi border. The suit, filed in Mississippi federal court, alleges the company is violating the Clean Air Act.

Data centers are energy-hungry, and the AI boom has accelerated their expansion at a pace that environmental regulators haven't kept up with. The lawsuit alleges that xAI's Memphis-area facility is generating harmful air pollution that disproportionately affects nearby communities of color - a pattern that critics of the tech industry's infrastructure buildout have been warning about for years.

This connects to a broader conversation about environmental concerns in AI infrastructure. The computing power required to run large models doesn't exist in a vacuum - it requires land, water, and energy, and the communities closest to that infrastructure often bear costs they never consented to. Regardless of how the lawsuit resolves, it adds legal pressure to an industry that has largely operated without this kind of scrutiny.

Read the full story →

If you use Gemini inside Chrome, this is a quality-of-life upgrade worth knowing about. Google has rolled out a feature called Skills - it lets you save any AI prompt as a reusable, one-click workflow that runs across whatever tab you're on. The rollout started April 14th and is targeting Mac, Windows, and ChromeOS users with Chrome set to English-US.

The idea is straightforward: if you find yourself typing the same prompt over and over - summarize this article, extract the key numbers, rewrite this in plain English - you can now save that as a Skill and apply it anywhere with a single click. Google has also shipped a set of premade Skills out of the box, including things like maximizing protein in recipes and summarizing YouTube videos.

It's not a revolutionary feature, but it's the kind of thing that makes AI genuinely more useful day-to-day rather than just impressive in demos. If you're building workflows on top of browser-based AI, check out our token calculator to keep an eye on what those repeated prompts are actually costing you.

Speaking of building fast - if you're using AI workflows to spin up landing pages or websites, 60sec.site is an AI website builder worth checking out. Describe what you need, and it generates a complete site in seconds - no design skills required.

Read the full story →

The answer is Embodied Reasoning! The 'ER' in Gemini Robotics-ER stands for Embodied Reasoning - a term that describes AI systems designed to reason about the physical world, including spatial awareness, task planning, and interpreting real-world environments. It's the distinction between an AI that can talk about picking something up versus one that can actually figure out how to do it.

The deepfake nudes story today raises a question I'm genuinely curious about: do you think AI companies are doing enough to prevent their tools from being weaponized like this, or does the responsibility lie elsewhere - with platforms, schools, or policymakers? Hit reply and tell me where you think accountability should sit. I read every response.

That's all for today. For more daily AI coverage, visit dailyinference.com - see you tomorrow!