• AI PlanetX
  • Posts
  • OpenAI Pauses GPT-6 Over Cyber Risk

OpenAI Pauses GPT-6 Over Cyber Risk

Anthropic Cuts Manual Claude Approvals

Author

Altiam Kabir
August 08, 2026

In partnership with

AI PlanetX

Welcome to another edition of AI PlanetX.

OpenAI slows Astra over cybersecurity risks; Claude Code defaults to safer auto mode as New Orleans tests AI for 911 calls.

Inside This Edition: 💎

  • Hottest AI News

  • Top AI & SaaS Tools

  • AI Tutorial: Build an AI Version of Yourself

  • Top AI & Tech News

  • AI Art Spotlight

  • Resource of the Day: Guide to AI Loops and Agent System

  • Featured AI Video

  • AI Meme of the Week

Hottest AI News

OpenAI

OpenAI Slows Astra Development Over Critical Cybersecurity Risk

OpenAI has slowed development of its upcoming Astra (GPT-6) model after tests showed major gains in agentic coding and cybersecurity, and it cannot rule out “Critical” capability under its Preparedness Framework.

Details:

  • At Critical level, a model could autonomously build zero-day exploits for hardened systems or devise and execute complex attacks from a high-level goal. GPT-5.6 Sol was rated High

  • OpenAI paused Astra activities that fail stronger rules, adding isolated testing, restricted network and tool permissions, tighter weight security, sandboxing and universal monitoring

  • Astra did not cause the Hugging Face breach. But recent incidents involving OpenAI, Anthropic and Meta show frontier agents are getting harder to contain during cyber testing

OpenAI says it disclosed this for transparency, will test Astra with government and safety groups, and wants advanced cyber models to strengthen defenders.

Turn AI into Your Income Engine

Ready to transform artificial intelligence from a buzzword into your personal revenue generator?

HubSpot’s groundbreaking guide "200+ AI-Powered Income Ideas" is your gateway to financial innovation in the digital age.

Inside you'll discover:

  • A curated collection of 200+ profitable opportunities spanning content creation, e-commerce, gaming, and emerging digital markets—each vetted for real-world potential

  • Step-by-step implementation guides designed for beginners, making AI accessible regardless of your technical background

  • Cutting-edge strategies aligned with current market trends, ensuring your ventures stay ahead of the curve

Download your guide today and unlock a future where artificial intelligence powers your success. Your next income stream is waiting.

Anthropic

Anthropic Makes Claude Code Auto Mode Default for Safer AI Development

Anthropic is making Claude Code’s Auto Mode the default, letting a safety classifier handle routine approvals while blocking risky actions. In tests, it caught dangerous commands much more often than human reviewers did.

Details:

  • From August 14, Pro, Max, and Team users get Auto Mode by default, while Enterprise stays opt-in. Claude works continuously and asks when actions look risky or irreversible

  • In a 1,053-user study, Anthropic says Auto Mode caught 89% of dangerous commands versus 13.6% for humans. Teams using it also created about 25% more pull requests

  • Anthropic says 720 prompt-injection attacks failed against Auto Mode. It also blocked data leaks and risky internal actions, but still recommends human review for production changes

Classifier checks are f-r-e-e, but longer runs can use more tokens. The bigger shift is developers moving from constant approval to AI oversight.

Top AI & SaaS Tools

  • Kollab (Life-time Deal): AI workspace that keeps project context and brand voice intact while helping teams create, repurpose, and publish content directly to social platforms from one place

  • Wispr Flow Notetaker: Improves meeting notes by accurately identifying speakers and preserving names, terminology, and context, making follow-ups and summaries more reliable [F-R-E-E to Try]

  • Hark Handoff: Claims to be the best internet-use model ever built for everyday web tasks like booking, shopping, and browsing

  • Prime Agent: Self-improving coding harness for long-running autonomous tasks, using programmatic tools, variable context, multi-agent messaging, and self-modifiable state

  • Kitesurf: Cloudflare’s agent-first browser built for AI automation, with strong isolation, lower CPU/memory use, and broad test coverage [F-R-E-E]

Stop Paying for 6 Tools. One AI Does It All.

Most e-commerce sellers juggle 6–8 tools and pay hundreds monthly to keep operations running. StoreClaw replaces the stack with one autonomous AI engine that monitors competitors, optimizes listings, automates marketing, and tracks profit 24/7. Connect your store and let AI handle the work — no prompts, no complex setup, no credit card required.

AI Tutorial

How to Create an AI Version of Yourself Using Pika

Use Pika to build an AI Self that matches your identity, tone, memory, appearance, and workflow. The better you set it up, the more helpful it is for messaging, content, brainstorming, reports, and daily tasks.

  1. Launch Pika

    Go to Pika or use the Pika iOS app. Sign up or log in to start creating your Pika Agent / AI Self.

  2. Start setup

    Begin the onboarding flow. Pika describes this as creating an AI Self or Agent that becomes a persistent, portable creative partner.

  3. Add personal context

    Give Pika the details that shape your AI Self.

    • Your taste

    • Your personality

    • Your preferences

    Example:

    “I create simple AI tutorials for beginners, creators, students, and small business owners.”

  4. Set communication style

    Tell your AI Self how it should sound when speaking or writing.

    • Clear and practical

    • Friendly but not too casual

    • Short, useful, and easy to understand

    Example:

    “Explain things simply, avoid robotic wording, and write like a helpful AI educator.”

  5. Create face, voice, and appearance

    Design how your Pika Agent looks and feels. Pika officially says your agent can have a face, voice, personality, and appearance that you create.

  6. Train it through conversation

    Talk to your AI Self and correct it when the response does not match you. The agent can remember your context, knowledge, preferences, and how you write across sessions.

    Example:

    “This sounds too formal. Make it more direct, simple, and creator-friendly.”

  7. Use it across workflow

    Use your AI Self for messaging, DMs, brainstorming, content, reports, tasks, and creative work.

Note: Correct it whenever the response feels off. Small feedback like “too formal” or “make it shorter” helps shape the AI Self over time.

Top AI & Tech News

  • New Orleans is testing AI to help answer and triage 911 calls, aiming to reduce the burden on human dispatchers and improve response times

  • Airbnb says AI is speeding up product development significantly, cutting feature delivery time by up to 60% and increasing shipped improvements by nearly 80%

  • Android Canary 2608 appears to add an AI-based lockscreen clock generator with prompt, image, and idea options, but the feature isn’t live yet

  • Dario Amodei would retreat to the in-office library and watch YouTube videos to calm down after conflicts with Sam Altman

AI Art Spotlight

Create Your Own Photo Like This Using the Prompt Below:

A premium iced caramel latte stands in the foreground on a thick white-veined marble pedestal, its transparent glass revealing layered espresso, milk, caramel ribbons, ice cubes, condensation, and glossy specular highlights. Warm sunlight enters from the upper left, producing soft directional shadows and amber subsurface glow through the drink while suspended roasted coffee beans create depth across the right side and lower left. The composition uses a restrained cream-and-brown palette, a spacious editorial background divided by a subtle geometric grid, and refined serif typography reading “The Art of Iced Perfection,” supported by minimalist navigation, rating, product, and call-to-action cards. Shallow depth of field isolates the glass while preserving the luxury beverage landing-page structure.

* **Style:** Photorealistic luxury beverage advertising, editorial web design
* **Key elements:** Iced caramel latte, coffee beans, marble pedestal, serif typography, interface cards
* **Lighting:** Warm diffused sunlight from upper left, soft shadows, 4000K amber-neutral palette
* **Camera:** Front-facing product shot, 50mm lens, eye-level angle, shallow DoF

Model: ChatGPT Images 2.0 / Nano Banana 2

Resource of the Day

The Complete Guide to AI Loops and Agent Systems

This guide explains how AI loops work and why they're becoming more effective than traditional prompting for many tasks. It covers the core building blocks of AI loops, when to use them, common pitfalls to avoid, and how to build self-improving workflows in Claude, ChatGPT, and other AI tools.

Featured AI Video

How to Run the Hermes Agent 100% F-R-E-E Using the Best Chinese Models

AI Meme of the Week

Every AI model trying to become a cybersecurity threat while Gemini is auditioning for an orchestra 😭