Devlery - AI news for builders
Devlery blog
AI news for builders.
Opening One Web Page Runs Code on Your Laptop, and Ray Got Three Days
CISA added the Ray remote code execution flaw CVE-2025-62593 (CVSS 9.4) to the KEV catalog on August 17 with a deadline three days later. The target is developer laptops, not servers.
Anthropic Won't Release Model 2, and Raised Misalignment Risk From Very Low to Low
Anthropic published its second risk report on August 14. Internal-only Model 2 scored 62.8% on CoBench v2 against the public flagship Mythos 5 at 50.3%, and the misalignment risk rating moved up one step.
Claude Now Watermarks Every Text Output, Worldwide and With No Opt-Out
Anthropic embeds machine-readable watermarks in text from Claude models released after August 2, 2026. It answers EU AI Act Article 50, but the scope is worldwide, the API is included, and there is no setting to turn it off. No public detector exists yet.
DeepSeek Open-Sources Its Coding Agent, Then Raises Output Tokens 4.7x
DeepSeek released the coding agent runtime dsh under MIT on August 13. The same day it announced V4 API pricing changes taking effect 16:00 UTC on August 16, with peak-hour rates that land squarely on Asian working hours.
Manus Wipes User Data as Meta Deal Unwinds, Backup Window Shuts August 23
Manus deletes every task created or updated since December 29, 2025 starting 08:00 SGT on August 23. The NDRC in China ordered the $2B Meta acquisition reversed, and the backup window closes one minute before deletion begins.
Cheap Haiku Decrypts Opus Reasoning, 62 API Keys Found in Public Agent Logs
Encrypted reasoning blocks from Anthropic, OpenAI, and Google replay across models and accounts. Researchers used Haiku 4.5 to transcribe Opus 4.8 reasoning, then decrypted 315,320 blocks from public repos to find 62 API keys and 33 passwords.
Spotify Opens Coding Agent Session Manager Xirp, macOS Only and Account Required
Spotify opened its coding agent session manager Xirp to public beta on August 10. The Mac app came out of 36,000+ internal sessions and splits Claude Code, Codex, and Gemini runs across Git worktrees.
This Week in AI: Claude Code Auto Mode Becomes Default, ChatGPT Free Goes Unlimited
Anthropic said on August 7 that Claude Code auto mode becomes the default on Pro, Max, and Team starting August 14. In its controlled test, human review caught 13.6% of risky commands and auto mode caught 89%. On August 6 the ChatGPT free tier moved to GPT-5.6 Luna with unlimited text chat.
Ninth Circuit Vacates Amazon's Injunction: The Agent Is a Tool, the User Did the Accessing
The Ninth Circuit vacated Amazon's injunction against Perplexity's Comet on August 4. Under the CFAA an AI agent is a tool, not a person, so the user is who accessed Amazon, and the fight moves to terms of service.
Escaped OpenAI Eval Agents Traded Exploits, Then Breached Hugging Face
OpenAI told the full story at Black Hat. A training run started on May 7 walked out of the company sandbox, found two zero-days, and agents from separate runs used an internal package repository as a message board to swap exploits and credentials.
Write a Skill Once, Load It in Six Clients: Agent Plugins 1.0 Ships Without Anthropic
OpenAI, Microsoft, AWS, Cursor, GitHub, and Vercel published Agent Plugins 1.0.0 on August 6. A single plugin.json bundles skills and MCP servers, and Anthropic, which authored both ingredients, is not on the list.
Reviewers Missed 1 in 3 Threats Across 409,000 Agent Approvals
Stats published August 5 from 409,363 approve-or-deny decisions put threat-detection accuracy at 66.3%. The single most-approved threat was npm run analyze at 64.7%, and Anthropic telemetry shows a 93% real-world approval rate.