Want SOC 2 compliance without the Security Theater?

Question 🤔 does your SOC 2 program feel like Security Theater? Just checking pointless boxes, not actually building security?

In an industry filled with security theater vendors, Oneleet is the only security-first compliance platform that provides an “all in one” solution for SOC 2.

We’ll build you a real-world Security Program, perform the Penetration Test, integrate with a 3rd Party Auditor, and provide the Compliance Software … all within one platform.

Schedule a demo for pricing

Today:

• Murati Leaves OpenAI

• OpenAI Restructuring and Updates

• Gemini and NotebookLM Updates and Features

• Microsoft AI Investment

• Meta Introduces Orion AR Glasses

BREAKING: Sam Altman STUNNED As Top Employees LEAVE OpenAI

Mira Murati, a key leader at OpenAI who briefly served as CEO when Sam Altman was temporarily removed, has announced she is leaving to explore new opportunities. She thanked Sam and Greg for their support and mentioned recent breakthroughs like speech-to-speech technology. 

Her departure continues a trend of important people leaving OpenAI to start new projects, leading some to question what's happening inside the company. Sam Altman thanked her for her contributions and said more details about the transition will come soon. 

WATCH THE VIDEO ON YOUTUBE

OPENAI

OpenAI is making significant changes to its business and technology. The company plans to restructure into a for-profit benefit corporation, moving away from control by its non-profit board to attract more investors. The non-profit arm will still exist and own a minority share. CEO Sam Altman will receive ownership stakes for the first time, potentially valuing the company at $150 billion after the change. This shift coincides with leadership changes, including the sudden departure of Chief Technology Officer Mira Murati and President Greg Brockman taking leave.

On the technology front, OpenAI is enhancing its Moderation API with a new model based on GPT-4o. This upgraded system is better at detecting harmful text and images, works in multiple languages, and provides more precise moderation decisions. It's available for free to developers, helping them build safer applications.

OpenAI is also launching the OpenAI Academy to invest in developers and organizations using artificial intelligence to tackle tough problems and boost economic growth, especially in low- and middle-income countries. The Academy will offer training, technical support, free access to OpenAI models, and community-building activities.

Additionally, the company is revamping Sora, its video-generating model first previewed in February. The new version aims to create longer, higher-quality video clips more quickly. Advanced voice features are also rolling out in the ChatGPT app for Plus and Team users, including custom instructions, memory functions, five new voices, and improved accents in over 50 languages.

GOOGLE

Google has announced major updates to its AI tools. NotebookLM, a platform that helps users understand their own materials, now supports uploading YouTube videos and audio files in addition to documents like PDFs. This enhancement allows users to summarize key concepts from videos, search through transcribed audio recordings, and create comprehensive study guides from class notes with a single click.

Two new AI models, Gemini-1.5-Pro-002 and Gemini-1.5-Flash-002, have been introduced. These models are faster, more efficient, and offer improved performance in areas like mathematics and code understanding. Prices for using these models have been reduced by over 50%, making them more accessible to developers. They can handle large amounts of text and understand images, enabling tasks like analyzing lengthy documents or processing complex code.

In other news, Google reportedly paid $2.7 billion to rehire AI expert Noam Shazeer, who had left the company after disagreements over AI development. Shazeer co-founded Character.AI and is now a leader in Google's Gemini AI project. Google's generative AI tools are currently used by over 2 million developers, with companies like GoTo Group creating AI-based voice assistants for their apps.

MICROSOFT

Microsoft CEO Satya Nadella announced major investments in Mexico and Brazil to boost artificial intelligence (AI), which allows computers to perform tasks that typically require human intelligence. 

In Mexico, Microsoft will invest $1.3 billion to enhance AI infrastructure and teach AI skills to 5 million people, making AI accessible to all. In Brazil, the company will spend 14.7 billion Reais to expand cloud and AI services and launch ConectAI, also aiming to train 5 million people. These efforts will help small businesses, improve internet access, tackle health and environmental issues, and ensure that the benefits of AI reach everyone.

META

At its 11th annual Connect conference, Meta CEO Mark Zuckerberg announced new products in mixed reality, artificial intelligence (AI), and wearable technology. Meta is launching the Quest 3S headset for $299.99 on October 15. This device combines the real world with virtual elements, letting users interact with digital objects in their physical space. 

The company also enhanced its AI assistant, Meta AI, now used by over 400 million people monthly, which can understand both images and text. Additionally, Meta introduced Orion, a prototype for lightweight glasses that display digital information over the real world, aiming to deepen human connections through technology.

🧠RESEARCH

Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Multimodal Models

Molmo is a new open-source model that understands images and text. They created a detailed dataset by having people describe images using speech. Molmo outperforms other open models and rivals top proprietary systems. They plan to release all their data, code, and model details soon.

HelloBench: Evaluating Long Text Generation Capabilities of Large Language Models

HelloBench is a new tests to measure how well large language models generate long texts. They found most models can't produce texts over 4,000 words and often repeat themselves or lose quality. They also developed HelloEval, a way to evaluate results that aligns closely with human judgments while saving time.

RACER: Rich Language-Guided Failure Recovery Policies for Imitation Learning

RACER is a new method that helps robots recover from mistakes using detailed language instructions. They created a system where a vision-language model acts as a supervisor, guiding a robot's actions. RACER outperforms existing methods in various tasks, both in simulations and real-world tests.

Imagine yourself: Tuning-Free Personalized Image Generation

"Imagine yourself" is a new model for personalized image creation that doesn't need individual tuning. It overcomes previous models' problems of limited image diversity and poor adaptation to prompts. Using new techniques, it produces high-quality, diverse images while keeping the person's identity and following text descriptions accurately.

Programming Every Example: Lifting Pre-training Data Quality like Experts at Scale

ProX is a new method that allows small language models to improve training data by programming adjustments for each example. This leads to better models than traditional data cleaning methods, saves computing power, and works across different model sizes and datasets. ProX will be shared openly.

🛠️TOP TOOLS

Notion - Search, generate, analyze, and chat—right inside Notion.

Magnific AI - Image Upscaler, Enhancer & Transformer

Nuts and Bolts - Simplify your prospecting process with powerful Google Sheets based AI tool. Simply input your target companies, and let copilot do the heavy lifting.

Blaze - Creates all the marketing content you need and posts it for you.

NodeLand - Transform notes into incredible mind maps using AI

📲SOCIAL MEDIA

🗞️MORE NEWS

• The Allen Institute for AI released Molmo, open-source AI models that outperform top private models on several tests. Molmo can understand images and uses much less data due to new training methods.

• Intel unveiled Xeon 6 processors with stronger cores and Gaudi 3 AI accelerators, which double AI performance and reduce costs. These new tools aim to make advanced AI systems more accessible to companies.

• Airtable launched new tools called App Library and HyperDB, turning its app-building platform into a powerful artificial intelligence system for businesses. These tools help companies use AI in their work, making it easier and more flexible.

• Notion launched an AI tool that lets you search, create, analyze, and chat within Notion. It finds answers from apps like Slack, writes documents in your style, examines files, and uses advanced AI like GPT-4.

• Duolingo launched AI-powered mini-games called Adventures and a Video Call feature that lets users practice languages by talking to an AI character. These new tools make language learning more interactive and realistic.

• Alibaba and Nvidia are collaborating to integrate Alibaba's AI language models into Nvidia's self-driving car platform. This partnership aims to enhance smart vehicles with improved voice assistants and intelligent in-car experiences.

• AI company Runway is giving up to $1 million to filmmakers who use its AI tools, aiming to speed up AI-made films. This follows a deal with Lionsgate to train AI on their movies.

• Nebius, an Amsterdam company formed after separating from Russia's Yandex, will invest over $1 billion in Europe by mid-2025 to build facilities supporting artificial intelligence, like new data centers and increased computing power.

• Director James Cameron joined Stability AI's board—the company behind the AI model Stable Diffusion. He believes that merging AI with computer-generated images is the future of filmmaking, opening new ways for artists to tell stories.