TODAY'S TOP STORIES — MAY 27, 2026
- China AI Travel Restrictions — Government approval now required for top Alibaba and DeepSeek researchers to travel abroad
- Gemini 3.5 Flash GA — Google's fastest frontier model now default in AI Mode Search for 1B+ users; beats Gemini 3.1 Pro on every agentic benchmark
- OpenAI Daybreak — GPT-5.5 + Codex Security cybersecurity initiative goes public, directly challenging Anthropic's Project Glasswing
- BingX Pre-IPO Futures — USDT perpetual contracts on OpenAI and Anthropic pre-IPO valuations now live, joining HTX and OKX
1. China Restricts Overseas Travel for AI Researchers at Alibaba and DeepSeek
Bloomberg reported on May 26 that Chinese government agencies have begun requiring prior approval before top AI professionals at Alibaba Group and DeepSeek can travel abroad. The restrictions target startup founders, researchers, and senior executives working on advanced AI — not ordinary employees — and represent a significant escalation from the prior regime, where staff had to report travel plans but did not always need formal sign-off before departing. Neither Alibaba nor DeepSeek responded to Bloomberg's requests for comment.
Beijing has maintained travel restrictions on individuals in state-adjacent sensitive sectors for years — nuclear scientists, senior executives at state-owned enterprises, university researchers with classified affiliations. Extending these controls into private-sector AI operations is the new development. Bloomberg had previously reported that some DeepSeek executives faced similar restrictions as far back as December 2025, and two co-founders of Chinese AI startup Manus were also reportedly barred from overseas travel earlier this year. The May 26 report represents a broader expansion across private-sector AI. Chinese regulators also blocked Meta's planned $2 billion acquisition of Manus earlier in 2026 — the consistent thread is technology and talent retention inside China.
Full coverage: China Blocks AI Researchers at Alibaba and DeepSeek from Traveling Abroad →
2. Gemini 3.5 Flash Reaches General Availability
Google launched Gemini 3.5 Flash at Google I/O on May 19 and it is now generally available across the Gemini API, Google AI Studio, Google Antigravity, Android Studio, and the Gemini app. It replaced the prior default model in AI Mode Search, which has surpassed one billion monthly users globally. Google CEO Sundar Pichai described the launch as the start of the "agentic Gemini era."
The key benchmarks: 76.2% on Terminal-Bench 2.1 (coding), 1656 Elo on GDPval-AA (agentic task performance), 83.6% on MCP Atlas (multi-tool coordination), and 84.2% on CharXiv Reasoning — all beating the previous Gemini 3.1 Pro, while running at 4x the output token speed of comparable frontier models. Pricing is $1.50 input / $9.00 output per million tokens with a 1M-token context window. That is 3x the price of the previous Gemini 3 Flash Preview, though still 25% below Gemini 3.1 Pro. Gemini 3.5 Pro, the larger sibling, is in internal use at Google with public rollout expected in June 2026.
Full review: Gemini 3.5 Flash — Benchmarks, Pricing, and How It Compares to GPT-5.5 and Claude →
3. OpenAI Launches Daybreak — Its Cybersecurity Answer to Anthropic's Glasswing
OpenAI launched Daybreak on May 11, 2026 — a cybersecurity initiative that combines GPT-5.5 models with Codex Security to continuously scan codebases for vulnerabilities, build threat models, validate attack paths in isolated environments, and propose patches for human review. The goal is to move security from a quarterly post-deployment audit into a continuous process running inside the development loop. Early partners include Cloudflare, Cisco, CrowdStrike, Oracle, and Zscaler. Any organization can request a scan at openai.com/daybreak.
Daybreak is OpenAI's direct response to Anthropic's Project Glasswing, which launched April 7 using the unreleased Claude Mythos model — stronger than Opus 4.7 on security tasks, but restricted to approximately 50 partner organizations including AWS, Apple, Cisco, Google, JPMorgan Chase, and Microsoft. Glasswing has a notable proof point: Mozilla used Mythos through Glasswing to find and patch 271 vulnerabilities in Firefox ahead of a major release. The key difference is access: Daybreak is open to any organization today; Glasswing is not.
Full breakdown: OpenAI Daybreak vs. Anthropic Project Glasswing — Side-by-Side Comparison →
4. BingX Launches OpenAI and Anthropic Pre-IPO Futures Contracts
BingX has added OpenAI and Anthropic to its Pre-IPO Trading Zone as USDT-margined perpetual futures contracts — traders can go long or short 24/7 without owning any equity, voting rights, or IPO allocation. Early reference prices were approximately $1,600 for OpenAI (implied valuation: $852B post-money after its March 2026 $122B raise) and $1,800 for Anthropic ($380B post-money after its February 2026 $30B raise). BingX is also running a 300,000 USDT airdrop campaign on the OpenAI contract through May 28.
The category is now standard across major crypto derivatives exchanges: HTX has both OpenAI and Anthropic live, Binance launched a SpaceX perpetual (SPCXUSDT) on May 21 with more planned, and OKX has announced OpenAI, Anthropic, and SpaceX contracts without yet giving a launch date. Bloomberg has reported Anthropic is targeting an October 2026 IPO; OpenAI is expected to follow in Q4 2026.
Full article: BingX Launches OpenAI and Anthropic Pre-IPO Futures Contracts →