Microsoft Bows Tools to Deal with Rogue AI Agents; Apple-New Products Next Four Fridays; Anthropic-5% Shares to Benefit Fund & Musk to Give Every US Kid a Share in SpaceX; Newest OpenAI and Google Models Easily Tricked into Making Non-consensual Nudes

Microsoft has rolled out tools to monitor and control rogue AI agents. Geekwire.com reports that the new features are part of Microsoft 365 E7…the company’s top-tier bundle of workplace software and AI tools, and its Project Perception security system. The new Microsoft tools come on the heels of a now steady flow of reports about AI agents breaking out of sandboxes and breaking into websites, stealing credentials, and other mischief. Microsoft CEO Satya Nadella wrote in a post that “The most trustworthy Super Intelligence system will not be the one with the model we trust most. It will be the one that enables us to trust the model the least.” Redmond plans to reveal more details at their Ignite conference November 17-19 in San Francisco.

Apple will be releasing new products on the next 4 Fridays. According to macrumors.com, the sold-out iPhone Duo will be released October 23rd (if you don’t have one ordered, expect to get one first quarter next year!). Then, of Friday, October 30th, the just-announced HomePod Mini 2 will drop. Rolling into November, the MacBook Pros with OLED displays and all the hardware upgrades will be out November 6th, followed by HomeView and Apple TV 4th generation on November 13th. All should have good availability on launch, with the exception mentioned above…the iPhone Duo folding phone…which is so hot that now Samsung says they are seeing a nice increase in folding phone sales, too.

Anthropic has said that…subject to shareholder approval…they would commit 5% of the company’s shares to public benefit fund. On the heels of that, Elon Musk is now saying he will give one share of SpaceX to every child in the US next year. Cnbc.com notes that there were 72 million kids under 18 as of July 1st, 2025. That would mean that Musk would give away some $10,5 billion worth of shares, based on SpaceX’s closing stock price as of last Friday.

Research from White Circle…a public benefit corporation backed by folks at Anthropic, DeepMind, Hugging Face, Google, and OpenAI has found that both Google’s  AI models and OpenAI’s can be easily tricked into generating non-consensual explicit content. Arstechnica.com reports that the AI models can basically be jail broken using just off-the-shelf tools…available for anyone to download. The research found that…to sum things up…“at a time when AI safety has never been under more scrutiny, the defenses leading labs have built are far weaker than they acknowledge.” Perhaps the worst thing found in the research is that the new models ChatGPT Images 2.5 and Nano Banana 2.1 are both significantly more breakable than their previous versions.” I would say we can look for the race between people making deepfakes and AI companies trying to stop them to be much like that between the police radar makers and radar detector makers…a constant battle where each side has the upper hand for a minute, then the other passes them up.

I’m Clark Reid and you’re ‘Technified’ for now.