Latest Software & Apps News
Anthropic Restricts Claude Mythos Amid Frontier Security Risks
Anthropic limited Claude Mythos to a small corporate cohort via Project Glasswing, offering $100M in credits for defensive vulnerability hunting. This contrasts with OpenAI's general release of GPT-5.5, dubbed 'Mythos-like hacking open to all.' The AI Security Institute confirmed the capabilities' real impact on collapsing vulnerability-to-exploitation timelines.
OpenAI Releases GPT-5.5 with Advanced Hacking Capabilities
OpenAI made GPT-5.5 generally available, enabling mythos-like hacking features that surpass human experts in finding and exploiting vulnerabilities. This follows Anthropic's cautious approach and has prompted government discussions on risks. Evaluations confirm its consequential security implications.
Trump Administration Considers Pre-Release AI Model Reviews
The White House is discussing executive orders for government oversight of advanced AI models, including pre-release vetting for high-risk systems capable of cyberattacks. This reconsiders abandoned Biden-era safety evaluations amid Anthropic's Mythos alarms. No finalized proposals yet, per NYT, WSJ, Axios reports.
Open-Weight AI Models Reshape Economics with Pricing Pressure
At least 10 major model releases this cycle from DeepSeek, Alibaba, Google, Z.ai, Moonshot challenge closed providers like Claude Opus on coding benchmarks at lower costs. Stanford AI Index documents the shift, forcing trade-offs on cost, portability, support. Organizations must now select viable open alternatives.
Apple App Store Sees 84% Surge in New AI-Generated Apps
Q1 2026 brought an 84% increase in new apps vs. 2025, driven by AI ease like 'vibe coding.' Apple is cracking down, removing several vibe coding apps to limit such creations. This highlights AI's role in rapid app development.
Google Launches Gemini 3.1 Flash TTS with Advanced Voice Control
Gemini 3.1 Flash TTS offers unprecedented control over accents, style, expression in text-to-speech. Part of broader AI model updates reshaping capabilities. Announced alongside Stanford's 2026 AI Index and Meta's Muse Spark.
Stanford Releases 2026 AI Index Report with 400+ Pages of Data
The comprehensive 400-page report analyzes AI's state, documenting open model surges and benchmarks. Covers developments across providers and economics. Essential reading for AI builders.
Meta Releases Muse Spark Multimodal Model for Product Integration
Meta's refactored AI lab debuted Muse Spark, designed for seamless integration with Meta products, with future Contemplating Mode for agent orchestration. Part of competitive model cycle.
Intel Highlights AI Compute Ecosystem at Computex 2026
Intel's CEO keynote on June 2 will showcase open platforms, partners, and progress from AI PCs to cloud, emphasizing silicon, software, systems for scaling AI. Demonstrates real-world momentum.
Cisco Acquires Astrix Security for AI Agent Permission Controls
Cisco bought Astrix Security for $400M to enhance software controlling AI agent permissions. Part of broader M&A in AI security amid rising concerns.
OWASP Unveils Plan to Eliminate Insecure Software at Scale
OWASP's strategic plan provides a roadmap tackling insecure software globally, urgent amid AI-driven vulnerability discoveries. Aims for a world without software insecurities.
U.S. Leads AI Security Investment but Lags in Deployment
Survey reveals highest global belief-to-deployment gap in AI security for U.S. workforce despite top investment intent. Highlights implementation challenges. Published May 5, 2026.