News
OpenAI strengthens GPT-5.6 against prompt injection attacks with internal AI red team
53+ min ago (482+ words) A new adversarial AI called GPT-Red helped slash prompt injection failures by 6x, hitting a 0.05% failure rate against its own toughest attacks OpenAI’s latest model family, GPT-5.6, launched on July 9 with a quiet but significant upgrade under the hood. The company…...
OpenAI's GPT-Live can check your flights, pull stock prices, and hold a conversation all at once
35+ min ago (177+ words) The new voice model delegates complex tasks to background systems like GPT-5.5, marking a fundamental shift in how conversational AI handles real-time multitasking. OpenAI just dropped what might be the most significant upgrade to ChatGPT Voice since the feature launched....
OpenAI raises ChatGPT custom instructions limit to 5,000 characters
48+ min ago (418+ words) Paid users on Pro, Enterprise, Business, and Education plans get more than triple the previous character allowance for personalized AI behavior OpenAI quietly handed its paying customers a meaningful upgrade this week. As of July 15, 2026, the custom instructions character limit…...
Grok Build open-sources code and resets usage limits for users
1+ hour, 29+ min ago (400+ words) xAI's coding agent goes fully open-source after privacy controversy, deleting all previously uploaded user data in the process xAI just did something unusual for a company in the AI arms race: it gave away the code. Grok Build, the company’s…...
SPACE cuts sandbox creation latency to 60 ms for Perplexity Computer, raising the bar for AI infrastructure
4+ hour, 57+ min ago (418+ words) Perplexity's autonomous agent platform runs complex workflows in isolated Linux environments, and the infrastructure powering it is absurdly fast Creating an isolated computing environment used to be the kind of thing that took seconds, sometimes minutes. SPACE, the infrastructure provider…...
Anthropic's Claude captures 9% of global generative AI traffic in June
5+ hour, 30+ min ago (318+ words) Claude's web traffic share jumped from 2.2% to 9.2% in six months, leaving Grok and Copilot well behind Six months ago, Claude was a footnote in the generative AI traffic race. Today, it holds roughly one in eleven visits across the entire…...
Centrifuge expands bug bounty program to cover V3.1 upgrade, offering $250K for critical findings
5+ hour, 59+ min ago (197+ words) The real-world asset tokenization protocol has logged 480 findings to date with zero exploits since its 2019 mainnet launch Centrifuge has expanded its bug bounty program to include its V3.1 upgrade, with a top payout of $250,000 for anyone who finds a critical vulnerability....
Microsoft patches record 570 vulnerabilities in single update as AI supercharges threat discovery
5+ hour, 47+ min ago (305+ words) The tech giant's July Patch Tuesday nearly tripled the previous record, with implications for crypto infrastructure running on Windows and Azure environments. Microsoft just fixed 570 security vulnerabilities in a single monthly update. To put that in perspective, the previous record…...
Unpatched Cursor vulnerability exposes users to code execution risk
7+ hour, 36+ min ago (293+ words) The AI-powered code editor used by more than half of Fortune 500 companies has faced a string of critical security flaws, raising questions about the safety of AI development tools Cursor, the AI-powered code editor developed by Anysphere, is facing renewed…...
X to open source entire codebase after security review, says Musk
8+ hour, 27+ min ago (383+ words) The move invites independent reviewers to verify what's actually running under the hood of the platform formerly known as Twitter Elon Musk announced on July 15 that X will open-source its entire codebase once a security vulnerability review is complete. The…...