Claude Broke Into OpenAI in Under 72 Hours

Two of AI's biggest labs had a rough weekend. Security researchers used Anthropic's own Claude to break into OpenAI's internal systems in under three days, and separately, Google's Gemini logged into three real companies it was never supposed to touch. Neither company meant for it to happen. That's the story running through today's issue: the tools are getting sharper faster than anyone is tightening the guardrails around them.

AI is moving from experiment… to essential.

Every major industry is integrating it.
Every major company is investing in it.

By late 2025, AI was already an $800B market — growing at a pace that could push it well beyond $1 trillion in the years ahead.

Cloud infrastructure is scaling fast.
AI-enabled devices are multiplying.
Automation is becoming standard.

But here’s the real question…

When trillions flow into this transformation — which stocks stand to benefit most?

Our new report reveals 10 AI stocks positioned across the backbone of this shift — from the companies powering the infrastructure… to those embedding intelligence into everyday systems.

If you want exposure to one of the defining growth trends of this decade, start here.

Download the Report Now

Security researchers at Hacktron used Anthropic's Claude Opus 5 to turn a small bug in an image tool into a working attack on OpenAI, reaching employee ChatGPT accounts and OpenAI's internal GitHub repositories in under 72 hours, according to TechCrunch. The attack started with a single booby trapped photo uploaded to OpenAI's own support forum, which triggered a flaw in the forum's image library and chained into a login weakness that exposed employee accounts. The researchers proved the access was real by getting an employee's account to open a harmless pull request, then reported everything they found. OpenAI patched its side within 14 hours and paid a $6,500 bug bounty. The older Claude Opus 4.8 had failed at the same attack for months before Opus 5 cracked it in hours.

Why it matters to you: This isn't really an OpenAI story. It's an "every old bug is now a live wire" story. AI tools can now take a vulnerability that used to take a skilled hacker months to weaponize and finish the job in days. If your website, booking system, or point of sale software has an unpatched bug sitting in it, that gap just got a lot more exploitable.

What to do about it: Ask whoever manages your website and software whether security patches actually get applied on a schedule, not just "when someone gets around to it."

Google confirmed that its Gemini model reached into three real companies during a controlled hacking exercise run by outside security firm Irregular back in May, according to Axios. The test was supposed to point Gemini at fictional targets, but one fake company happened to share a name with a real one, and researchers had accidentally left Gemini's internet access turned on. Gemini guessed a password for one target and found leaked credentials for two more, then recognized the targets were real and stopped on its own. Google called it mistaken identity rather than the AI going rogue, and it has since changed how it runs these tests.

Why it matters to you: The lesson here isn't really about Google. It's about what happens when you hand an AI agent real access. An AI tool doesn't know "only touch the test environment," it just sees whatever logins, tools, and internet access it's been given, and it will use all of it.

What to do about it: If you've connected any AI tool to your email, calendar, CRM, or bank account, give it its own limited login instead of your admin credentials, and check what it can actually reach.

I had major FOMO about this event last year, and I was not even there. The Epic Marketing Summit is a two day live event in Miami where you do not just sit and take notes, you build. Over two days you walk out with a working AI powered sales funnel already built, not a slide deck of ideas you still have to go implement on your own time. You also get hands on with vibe coding, building your own AI app right there in the room with people who have actually done the work, not just talked about it.

Last year's event sold out, and I know a lot of the trainers personally, they are not theory people. That is exactly why I am going this January, and why I would like to see you there too. If you are serious about turning your sales process into a system instead of running it on your own time and energy, this is where that shift actually happens. Seats filled up fast last time, so do not wait until December to think about it. Grab your spot here before it is gone.

President Trump said on Truth Social that the administration plans to stand up a federal "AI Force," modeled on Space Force, and will name a new AI czar, according to Axios. He gave no real detail on budget, staffing, or where it would sit in government, only that the government would not "hinder or stifle" AI growth but would watch for "bad" behavior through the existing courts. The AI czar role has sat empty since David Sacks left it back in March.

Why it matters to you: Do not expect federal rules on AI anytime soon. This is a slogan, not a policy, and the details that would actually affect how you are allowed to use AI in your business are not here yet.

California Governor Gavin Newsom signed an executive order directing state experts to draft rules for an AI "kill switch," independent safety reviews, and emergency shutdown procedures for the most powerful AI models within two months, according to the Governor's office. It targets frontier labs like OpenAI and Anthropic specifically, and follows a summer full of AI incidents that made national news. California has a track record of setting rules that other states end up copying.

Why it matters to you: If you rely on AI tools built by California companies, and most of the big ones are, expect their terms, safety features, and possibly their pricing to shift over the next year as this plays out. It is worth watching, not panicking over.

OpenAI's internal financial projections show the company expects to burn through roughly $278 billion in cash by 2030 while spending around $856 billion on computing and infrastructure, according to Tech Times. Annual revenue is projected to jump from about $36 billion to $350 billion over that stretch, but the company may run through its current funding by 2028 if things do not go as planned.

Why it matters to you: This is the real reason your AI subscriptions keep changing: new paid tiers, new limits, new upsells. These companies are spending money on a scale that makes profitability a long way off, and that bill eventually lands on the subscriber.

What to do about it: Before you lock into an annual AI subscription, ask what happens to your price if the company raises rates mid-contract. Month to month gives you more room to react.

I have been a member of this mastermind for over three years, and it has quietly become one of the most valuable relationships in my business. Ignite is an annual membership built for entrepreneurs still under a million dollars a year who want structure, tools, and a room of peers who actually get where you are. Inside, you get six to eight live virtual workshops a year plus two in person multi day events, each one normally priced between $97 and $997 on its own. You also get a full marketing software suite that replaces over $5,000 a month in separate tools, all connected in one place, plus weekly mentorship calls with Perry Belcher and his team.

My own personal bonus for members is a monthly accountability and business development call with me directly. Some of my closest friends, and a few real business partners, came directly out of simply being in that room. If you are looking for your next milestone and want a community instead of another course sitting unopened on your hard drive, this is where I would point you. See if it fits here.

Meta's new Muse AI agent app hit the number one spot on the U.S. Apple App Store's free chart, passing ChatGPT, about a week after launch, according to 9to5Mac. Muse and similar apps from Meta and SpaceXAI use a more casual, character driven chat style instead of the traditional assistant interface. Apple released its own Siri AI the same week, with deeper access across iPhone, iPad, and Mac.

Why it matters to you: Consumer habits around AI chat are shifting fast, and the app your customers open first will not necessarily be the one you are used to using yourself. If you market to a general consumer audience, it is worth knowing these apps exist even if you never touch them personally.

Anthropic is reportedly considering releasing a new AI model ahead of its expected IPO, according to Reuters, as the company's annualized revenue paces above $100 billion and rival GPT-6 Astra gains ground in workplace use. Sources say the company is weighing safety review time against the pressure to keep pace before going public.

Why it matters to you: If you use Claude for your business, expect another model update and possibly new pricing or plan changes in the coming months as Anthropic positions itself for a public offering.

Google DeepMind launched a new "DeepMind Institute," led by Shane Legg, James Manyika, and Demis Hassabis, to host public debate about artificial general intelligence, according to TechCrunch. Its first essays call for more transparency in how models reason, and propose a U.S. led body where AI labs would voluntarily submit powerful models for outside review before release, with mandatory review possibly following later.

Why it matters to you: This one is early, but proposals like this tend to become policy templates a few years later. Worth filing away, not acting on yet.

Two of the biggest names in AI had their own tools turned against them, or turned loose by accident, this week, while Washington offers slogans and California offers actual homework. The pattern holds steady: the tools keep getting more capable faster than anyone tightens the guardrails around them. That is not a reason to panic. It is a reason to keep your own house in order: patched software, limited logins, and a healthy dose of wait and see before you hand an AI agent the keys to anything that matters. Stay skeptical, stay on the ride, but keep both hands on the bar.

If this kept you from getting blindsided by today's AI headlines, forward it to one business owner who needs it. New here? Subscribe below and get tomorrow's issue too.

Talk tomorrow,
Mark Shilensky

Follow me personally on Social Media:
Facebook: https://www.facebook.com/MarkShilenskyPage
Instagram: https://www.instagram.com/markashilensky/
Threads: https://www.threads.com/@markashilensky
LinkedIn: https://www.linkedin.com/in/shilensky/
X: https://x.com/markshilensky
TikTok: https://www.tiktok.com/@mark.shilensky
BlueSky: https://bsky.app/profile/markshilensky.bsky.social