🛡️ Claude Mythos Found 10,000 Security Flaws in 30 Days

📊 ChatGPT Lands in Microsoft PowerPoint

🧠 Mira Murati’s lab built AI that responds like humans

🌶️ OpenAI’s chip deal hit an $18B financing challenge

🛒 Alibaba is bringing Qwen AI into Taobao shopping

🖥️ Perplexity launched autonomous AI agents for Mac

🖱️ OpenAI dropped a Codex Chrome extension

🪓 Cloudflare cut 20% of staff as AI boosted internal productivity

Average growth across 2,000 companies

Growth by revenue band 

AI-led vs AI-enhanced. Who performed better? 

Claude Mythos is Anthropic’s AI-driven cybersecurity platform built for automated threat detection and analysis 

The company says the system identified over 10,000 vulnerabilities in 30 days across enterprise software environments 

Mythos uses autonomous AI agents to continuously scan codebases, infrastructure, dependencies, and configurations 

Findings included exposed credentials, outdated libraries, dependency risks, insecure configurations, and exploitable bugs 

The platform can validate exploitability, prioritize risks, and suggest fixes automatically

Anthropic says Mythos is designed for enterprise security teams managing increasingly large and complex software stacks

Move comes as AI labs race to build autonomous cyber defense systems for governments and enterprises 

PowerPoint’s upgraded Copilot now uses OpenAI models to generate complete presentations from prompts, documents, or meeting notes 

Users can automatically create slides, speaker notes, charts, summaries, and visuals using AI

AI tools can rewrite slides, change tone, shorten content, and redesign layouts instantly

Copilot can pull information from Word, Excel, Teams meetings, emails, and company files to build presentations faster 

Microsoft says the goal is to reduce repetitive presentation work and speed up enterprise productivity 

Features are rolling out to Microsoft 365 Copilot users globally

OpenAI’s AI chip project with Broadcom reportedly stalled after Broadcom requested Microsoft commit to buying 40% of the first “Jalapeno” chips. The custom hardware is meant to reduce OpenAI’s inference costs, but without Microsoft’s backing, the company may need new partners or financing terms.

Meta is reportedly developing an AI assistant called Hatch that can perform tasks autonomously instead of simply answering questions. The company is also preparing AI-powered Instagram shopping tools to help users discover and purchase products conversationally before year-end.

Alibaba plans to connect Qwen AI directly with Taobao and Tmall, enabling users to browse and purchase products through conversation. The system will support over 4 billion listings alongside delivery tracking, virtual try-ons, after-sales help, and price monitoring.

Anthropic is reportedly considering raising up to $50 billion at a valuation approaching $1 trillion. Investors are attracted by rapid revenue growth, though the capital is also needed to fund the enormous compute and infrastructure costs of frontier AI.

Perplexity launched its new Mac app featuring “Personal Computer,” enabling AI agents to interact with local files, Mac apps, and the web. The system shifts AI workflows from cloud-only experiences toward persistent desktop automation running directly on users’ devices.

OpenAI released a Chrome extension for Codex that automates repetitive browser tasks across tabs and websites. The tool combines browser controls, plugins, and coding capabilities to handle workflows involving dashboards, research, and logged-in web services.

Cloudflare is laying off around 1,100 employees despite record revenue growth, saying AI-driven productivity gains reduced the need for many support roles. The company claims AI tools have dramatically accelerated software development and operational efficiency.