“Artificial intelligence should not be conflated with real human intelligence.”

— Vint Cerf

Welcome to another edition of the Be10x AI Pulse,

From safety debates to enterprise adoption, this week's headlines show that making AI work is harder than building it. While Washington was busy deciding whether Anthropic's most powerful model was too dangerous to use, Microsoft and Meta were busy proving that building the AI is the easy part; actually making it work is another story entirely. Meanwhile, Microsoft is throwing $2.5 billion and 6,000 people at a different problem: getting companies to actually use AI, not just buy it.

From safety showdowns to billion-dollar pivots to CEOs owning up to miscalculations, this week shows the AI race isn't just about who builds the smartest model, it's about who can actually make it work.

Let’s jump right in!

Anthropic's flagship AI model, Claude Fable 5, is up and running again for everyone, everywhere, starting July 1 and it's a bit of a comeback story. Back in mid-June, the US government abruptly ordered Anthropic to cut off Fable 5 (and its even more powerful, more restricted sibling, Mythos 5) for any non-American user after Amazon's security researchers found a clever trick, called a "jailbreak", that could coax the AI into revealing or even writing up ways to exploit software weaknesses. Since Anthropic had no quick way to check every single user's nationality, it played it safe and switched the models off for absolutely everyone, sparking plenty of frustration in the process.

Anthropic, meanwhile, built a smarter safety filter that now catches that specific jailbreak trick over 99% of the time, automatically rerouting any blocked request to a less powerful model instead.

So as of this week, Fable 5 is fully back on Claude's apps and platforms worldwide (with some temporary usage caps easing in over the next few days), while Mythos 5 remains available only to a select group of trusted US organisations for now.

With enterprise AI adoption lagging behind the hype, Microsoft is embedding 6,000 engineers and consultants directly inside client companies following similar moves from Amazon, OpenAI, and Anthropic. Microsoft is putting real money behind a simple problem: lots of companies have bought into AI, but far fewer know how to actually use it well. The tech giant announced a $2.5 billion investment to launch a new division called Microsoft Frontier Co., which will place roughly 6,000 engineers, consultants, and sales staff directly inside client organisations to help them design, deploy, and manage AI systems, an approach the industry calls "forward-deployed engineering".

The move comes as Microsoft's rivals make similar bets. Amazon just committed $1 billion to its own AI rollout effort, while OpenAI and Anthropic set up comparable customer-deployment teams earlier this year.

Products like Microsoft 365 Copilot have drawn interest, but getting businesses to fully adopt AI in their day-to-day work has proven slower than expected, which is exactly the gap this new unit is meant to close.

Mark Zuckerberg gave Meta employees a rare dose of candour this week, admitting at an internal town hall that the company's massive AI-focused restructuring hasn't delivered what leadership expected. Earlier this year, Meta cut 8,000 jobs, roughly 10% of its global workforce, and shifted another 7,000 staffers onto AI teams, all to help fund an AI infrastructure buildout expected to hit $145 billion in 2026.

Zuckerberg told staff that top executives had "miscalculated" the timing of those changes and that AI agents, the tools meant to justify much of this investment, simply haven't advanced as fast as the company hoped. He pointed out that "the trajectory of agentic development over at least the last four months hasn't really accelerated in the way that we expected", despite Meta's early excitement about tools like Anthropic's Claude Code. The 8,000 layoffs have already stirred pushback and morale issues inside the company, but Zuckerberg struck an optimistic note anyway, telling employees he expects Meta's AI bets to start paying off in a more meaningful way within the next three to six months.

Rocket AI is a no-code website builder that lets you create professional websites without writing a single line of code. It offers access to 25,000+ customisable templates that can be edited using a simple visual interface; just update the text, swap the images, and tailor the design to any business or industry. Once your site is ready, you can deploy it with a single click, making it an excellent tool for freelancers, agencies, entrepreneurs, and anyone looking to build client-ready websites quickly.

Google Opal is a free AI workflow builder that lets you create powerful automations using simple natural language instead of coding. Just describe the task you want to automate, and Opal intelligently builds multi-step workflows by selecting the right tools for research, web searches, and other actions. With its intuitive drag-and-drop interface and ability to remember your preferences over time, it's an excellent tool for building smarter AI automations for personal use or business applications.

This week, we found a few interesting things for you to check out:

We love crafting every week’s issue for you, and would only want to serve the best content to you — so, if you could take out 1-minute to tell us how we did, it would help us improve the Be10x AI Pulse experience for you! :D

Newsletter Survey (1-min)

See you later,
The Be10x Editorial Team