AI News Of The Week (22nd May, 2026)
TLDR: Key AI Developments This Week
Google used I/O 2026 to launch Gemini 3.5 Flash and Gemini Omni, expand agentic features across Search and the Gemini app, and preview new AI-powered eyewear. OpenAI announced its first Applied AI Lab outside the US in Singapore with a S$300 million commitment. Anthropic briefed the Financial Stability Board on cyber vulnerabilities identified by Mythos, elevating frontier-model risk to a global financial stability concern. Washington moved closer to pre-release AI oversight with a new executive order framework, while Microsoft entered early talks to rent servers powered by its own AI chips to Anthropic.
Google Turns I/O Into a Full-Scale Push for the Agentic Gemini Era
On May 19, Google used I/O 2026 to launch Gemini 3.5 Flash and Gemini Omni, expand agentic features across Search and the Gemini app, and preview new AI-powered eyewear. The broader signal is that Google no longer wants Gemini to feel like a separate assistant; it wants it acting as a layer across search, browsing, creative tools, and wearables.
OpenAI Picks Singapore for Its First Applied AI Lab Outside the United States
On May 20, OpenAI and Singapore announced that the company will open its first Applied AI Lab outside the U.S., grow to around 200 local roles, and commit more than S$300 million to the city-state. That makes this week's move look less like a normal regional office opening and more like OpenAI planting a serious deployment and talent hub in Asia.
Anthropic Takes Mythos From a Tech Story to a Financial-Stability Story
On May 18, Reuters reported that Anthropic was set to brief the Financial Stability Board on cyber vulnerabilities identified by Mythos, after a request tied to Bank of England Governor Andrew Bailey. That is a notable shift, because it means frontier-model cyber risk is now being treated not just as an AI safety concern but as a potential issue for the global financial system.
Washington Moves Closer to Pre-Release Oversight for Frontier Models
On May 20, Reuters reported that President Donald Trump was expected to sign an executive order creating a voluntary framework under which AI developers would give the U.S. government access to covered models 90 days before public release and provide early access to critical infrastructure providers. The fight here is no longer about whether AI oversight is coming, but about how much visibility governments should have before the strongest systems go live.
Anthropic May Become a Customer for Microsoft's In-House AI Chips
On May 21, Reuters reported that Anthropic was in early talks to rent servers powered by Microsoft-designed AI chips, a potential win for Microsoft's Maia effort and another sign that Anthropic is shopping broadly for compute beyond Nvidia's ecosystem. It also suggests Microsoft is steadily building a future where its AI stack depends less exclusively on OpenAI.
Google DeepMind's Contextual AI Deal Shows the Talent War Is Getting More Creative
Also on May 19, Reuters reported that Google DeepMind agreed to hire more than 20 researchers from Contextual AI and license its technology in a deal worth roughly $80 million to $90 million. The structure matters because it reinforces how big AI labs are using licensing-and-hiring deals to capture talent and tech without pursuing full acquisitions that would face heavier antitrust scrutiny.
OpenAI Turns National AI Adoption Into a Product Strategy in Malta
On May 16, Reuters reported that OpenAI signed a deal with Malta to give all residents access to ChatGPT Plus for one year after they complete a free AI-use course, with Maltese citizens abroad also eligible. That makes the story more than a promotional partnership: it is one of the clearest examples yet of a government treating mass AI access and literacy as a national program rather than a private consumer subscription.
