- The Shift
- Posts
- Anthropic's Most Dangerous Model Exposed
Anthropic's Most Dangerous Model Exposed
Plus, ❓ The feud between Altman and Amodei is shaping every major AI decision, How to scale compliance without adding headcount using Delve, and more!
In partnership with
Welcome back to The Shift. Let’s get straight to what matters in AI today…
Today we have:
🔓 Anthropic Accidentally Leaked Its Most Powerful Model and It's Bigger Than Opus
❓How to Scale Compliance Without Adding Headcount Using Delve
⚔️ The Feud Between Altman and Amodei Is Shaping Every Major AI Decision
🔨Tools and Shifts you cannot miss
🔓 Anthropic Accidentally Leaked Its Most Powerful Model and It's Bigger Than Opus
A CMS misconfiguration exposed nearly 3,000 unpublished assets from Anthropic's blog, including a draft announcing Claude Capybara (also called Mythos), a new tier above Opus. The draft says Anthropic wants to "act with extra caution and understand the risks it poses, even beyond what we learn in our own testing." Too powerful and too expensive to release.
The Decode:
1. Capybara is a new tier, not an update - The leaked draft describes Capybara as "larger and more intelligent than our Opus models." It scores dramatically higher than Opus 4.6 on coding, reasoning, and cybersecurity. Anthropic confirmed it's real and is being tested with early access customers.
2. Cybersecurity Implications Spooked Markets - The draft warns Mythos is "far ahead of any other AI model in cyber capabilities." Cybersecurity stocks dropped 3-7% on Friday. Anthropic plans to release it first to cyber defenders before broader availability.
3. It's expensive, and that's a problem - Anthropic says the model is very expensive to serve and will be expensive for customers. Meanwhile, users on $100-$200 Max plans are already hitting rate limits within an hour during normal work. Compute demand is outpacing what Anthropic can afford to deliver.
4. The Leak Itself Is Embarrassing - Nearly 3,000 assets were publicly accessible due to default CMS settings. Security researchers found it independently. The cache also included details of a private CEO retreat in the UK with Dario Amodei. Anthropic called it human error.
Anthropic is building models that cost dramatically more to run while struggling to serve current ones reliably. The gap between what these models can do and what users can affordably access keeps widening. Reliability and pricing will define who actually benefits from the next generation of AI.
Together with Attio
Attio is the AI CRM for modern teams.
Connect your email and calendar, and Attio instantly builds your CRM. Every contact, every company, every conversation, all organized in one place.
Then Ask Attio anything:
Prep for meetings in seconds with full context from across your business
Know what’s happening across your entire pipeline instantly
Spot deals going sideways before they do
No more digging and no more data entry. Just answers.
❓How to Scale Compliance Without Adding Headcount Using Delve
Most teams hit a compliance wall when they grow. The work multiplies, the team doesn't. Here's how Delve removes that ceiling.
Step 1: Connect your stack Link AWS, GitHub, and your existing tools. Delve auto-collects evidence across your infrastructure from day one. No manual screenshots, no assigned compliance owner babysitting integrations.
Step 2: Let AI Copilot run in the background Delve monitors your compliance posture continuously, flagging gaps, organizing evidence, and keeping everything audit-ready without a dedicated hire.
Step 3: Automate security questionnaires As you land bigger customers, security reviews come with them. Delve fills them out automatically using your existing compliance data. Your engineering team never gets pulled in.
Step 4: Migrate from your current tool for free Already on another SOC 2 platform? Delve handles the full migration. Zero-touch. No disruption. No starting over.
Step 5: Grow into enterprise without growing your compliance team Audits stay clean. Reviews get handled. Headcount stays flat.
micro1 scaled compliance without adding a single hire. 11x closed enterprise deals faster. Bland unlocked $500K ARR in 7 days after switching.
Migrate now and get a $2,000 VISA card straight to your inbox upon onboarding. You can book a demo here.
⚔️ The Feud Between Altman and Amodei Is Shaping Every Major AI Decision
A WSJ investigation traces the rivalry back to a 2020 shouting match at OpenAI. A decade later, their disagreements over safety, speed, and control now drive every major decision both companies make.
The Decode:
It Started With Broken Promises - Dario Amodei nearly quit when Greg Brockman proposed selling AGI access to Russia and China, then Sam Altman promised him authority but quietly gave others the power to fire him, so he left in 2021.
The Split Was Philosophical - Dario Amodei believed you could scale fast and build safely, while Sam Altman chose speed and market dominance, and instead of resolving it, they built two companies around opposite answers.
The Feud Is Playing Out Publicly - Anthropic ran Super Bowl ads targeting OpenAI and Sam Altman called them "clearly dishonest," then Anthropic demanded Pentagon weapons safeguards while OpenAI signed without them and Defense Secretary Pete Hegseth threatened Anthropic with supply chain designation.
Privately It's Even Worse - Dario Amodei compared the Sam Altman and Elon Musk lawsuit to "Hitler vs Stalin," called Greg Brockman's $25M pro-Trump donation "evil," and likened OpenAI to tobacco companies with both now valued above $300B.
The question Dario Amodei left OpenAI to answer, can you build powerful AI and keep it safe and accessible, is now his biggest unsolved problem as the leaked Mythos model is dramatically ahead but too expensive to serve.
Together with 1440 Media
Tired of news that feels like noise?
Every day, 4.5 million readers turn to 1440 for their factual news fix. We sift through 100+ sources to bring you a complete summary of politics, global events, business, and culture — all in a brief 5-minute email. No spin. No slant. Just clarity.
🔨AI Tools for the Shift
⌨️ Clico - Bring AI directly into any webpage to answer questions, summarize content, and draft replies without switching tabs.
🌿 PollenTracker - Get a simple yes or no answer on whether it’s safe to go out based on pollen, air quality, and weather.
🎨 Oniichan - Create manga pages, anime characters, and dynamic panels from text with consistent art and no drawing skills needed.
🎭 ViralOps - Generate UGC-style ads at scale using AI actors or custom avatars without studios or production costs.
🎥 Swap Me In - Instantly swap faces into any video using AI with seamless results and no editing skills required.
🚀Quick Shifts
🧠 Bluesky introduced Attie, an AI-powered app that lets users create custom social feeds using natural language, enabling personalized algorithms and future no-code app building on its open AT Protocol.
🎵 Suno released v5.5 of its AI music model, introducing voice cloning, personalized taste learning, and custom model training to give users deeper control over how their music is created.
🧠 A Stanford study found AI chatbots often give overly agreeable, “sycophantic” advice, increasing user dependence and reinforcing harmful behavior, raising concerns about relying on AI for personal or emotional guidance.
🚪 xAI’s last remaining co-founders, Manuel Kroiss and Ross Nordeen, have reportedly left the company, completing a full leadership turnover as Elon Musk restructures and rebuilds the AI startup.
📈 Anthropic’s Claude is seeing a surge in paid subscribers, with consumer sign-ups and returning users rising sharply, driven by product launches, marketing campaigns, and increased public attention.
🧩 Prompt of the Day
Technological Innovations to Watch
Identify emerging technologies that can reshape your category and unlock new competitive advantages.
Paste the prompt: Drop this into ChatGPT and fill in your segment.
Prompt to paste
Research and summarize key technologies transforming the [Insert segment or industry] space. Include:
Technology Name: [Insert technology or trend]
Applications: [List practical use cases or implementations]
Relevance to Brand: [Explain how this technology can impact or benefit the brand]
Use Case: Leverage emerging tech to gain a competitive advantage and future-proof strategy.
That’s all for today’s edition see you tomorrow as we track down and get you all that matters in the daily AI Shift!
If you loved this edition let us know how much:
How good and useful was today's edition |
Forward it to your pal to give them a daily dose of the shift so they can 👇
Reply