Hi Synapticians!

OpenAI just launched GPT-5, and it's a big deal for ChatGPT. The main news? This new model is now available for free to all users, not just paid subscribers. It's the first time OpenAI is giving access to an advanced reasoning model without charging for it.

GPT-5 combines two approaches that were separate before: it can think deeply about complex problems while staying fast for simple questions. The system automatically chooses the best way to respond based on your request. No more choosing between different modes or settings.

What actually changes: no more mental load trying to figure out which model to use. Before, you had to pick between 4o for speed or o3 for complex problems. Now, GPT-5 decides on its own whether to respond quickly or take its time to think. The system adapts automatically to your question. Plus, it hallucinates way less than before. It makes up incorrect info in only 4.8% of cases, compared to over 20% for previous models.

For those who pay, there are perks. Plus subscribers at 20 bucks a month get higher usage limits. Pro subscribers at 200 bucks a month get unlimited access and can use GPT-5 Pro, a version that takes more time to think and give better answers.

Sam Altman talks about an important step toward AGI, but in reality, GPT-5 is just slightly better than the competition on most tests. Anthropic's Claude and Google's Gemini remain very close in performance. The real difference will probably be in the everyday user experience rather than technical benchmarks.

Here’s the rest of the news about AI today:

• A federal study identified 139 new methods to break top AI systems, but findings were allegedly withheld due to political pressure

• A hidden prompt in a Google Doc allowed ChatGPT to access a victim’s Google Drive without interaction

• Google is rolling out a new 'Guided Learning' feature in its Gemini app, designed to help users break down complex topics step by step

Top AI news

1. OpenAI releases GPT-5
OpenAI launched GPT-5 on August 7, 2025, now free for all ChatGPT users. This "unified" model transforms ChatGPT into an AI agent capable of building apps and completing complex tasks. GPT-5 slightly outperforms competitors with 74.9% on coding tasks and reduces hallucinations to 4.8%. Read online 🕶️

2. U.S. Government Suppressed AI Vulnerability Study
A federal study identified 139 new methods to break top AI systems, but findings were allegedly withheld due to political pressure. This comes as new federal rules now require the kind of testing the study performed. The case raises concerns about transparency and political interference in AI safety regulation. Read online 🕶️

3. Invisible Prompt in Google Doc Exposes Google Drive via ChatGPT
A hidden prompt in a Google Doc allowed ChatGPT to access a victim’s Google Drive without interaction, bypassing normal security. This highlights a major data privacy risk and shows how AI can be manipulated to breach cloud-based services. Urgent security upgrades are needed as AI tools become more embedded in daily apps. Read online 🕶️

4. Google Gemini's Guided Learning Feature
Google is rolling out a new 'Guided Learning' feature in its Gemini app, designed to help users break down complex topics step by step. This innovative approach utilizes a dialogue of questions and explanations to deepen user understanding. The feature aims to make learning more accessible and effective by providing interactive guidance through intricate subjects. It represents a significant advancement in AI-powered educational tools, focusing on active engagement for better comprehension and retention. Read online 🕶️

Meme of the Day

Stay Connected

Feel free to contact us with any feedback or suggestions; we’d love to hear from you !