Google I/O Reshapes Search and AI Agents, Trump Halts AI Executive Order, Apple Opens iOS to Third-Party AI — AI News Briefing

🗞️ AI News Briefing — May 22, 2026 (06:00 CST)

Top 7 Stories

1. Google I/O 2026: Gemini 3.5, Personal AI Agents, and the Biggest Search Redesign in 25 Years

Google I/O 2026 has delivered the company’s most ambitious AI announcement slate to date, fundamentally reshaping Google Search for the first time in 25 years. At the developer keynote, CEO Sundar Pichai unveiled Gemini 3.5 — a new family of models built specifically for agentic workloads that can sustain autonomous sessions for several hours and run complex coding pipelines independently. The company also announced Gemini 3.5 Flash, which scores 55 on the Artificial Analysis Intelligence Index, nine points above its predecessor Gemini 3 Flash, and achieved an Elo score of 1,656 on agentic tasks — a massive leap from Gemini 3 Flash’s 1,204.

The personal AI agent announcements represent Google’s most direct challenge to OpenAI and Anthropic in the agent race. Google’s new AI agent can draft emails, monitor inboxes, and eventually execute financial transactions on behalf of users. The agent integrates deeply with Google’s ecosystem, including the newly announced AI-powered Google Shopping experience with a Universal Cart feature. Google also unveiled new AI coding tools designed to compete directly with Anthropic’s Claude Code and OpenAI’s Codex, signaling that all three companies are now in a three-way race to dominate the developer tooling market.

The Search redesign is the most significant change to Google’s core product since its founding. Powered by AI, the new Search experience moves beyond traditional link results toward a more conversational, action-oriented interface. The New York Times characterized the change as “Powered by A.I., Google Changes Its Search Box for the First Time in 25 Years.” Google also announced a partnership to release smart glasses, expanding its AI ambitions into hardware. Pichai estimated that companies shifting 80 percent of workloads to a mix of Gemini 3.5 Flash and Pro could save over a billion dollars annually, though independent analysis by The Decoder notes that Gemini 3.5 Flash follows the industry-wide trend of making newer AI models significantly more expensive per token.

2. Apple Opens Apple Intelligence to Third-Party AI Models in iOS 27

In a major shift for Apple’s AI strategy, iOS 27 will allow users to choose between Anthropic, Google, and OpenAI models to power Apple Intelligence features. Sherwood News and Mashable both confirmed the development, reporting that the change will let users swap out Apple’s built-in models for third-party alternatives. This represents a significant departure from Apple’s historically walled-garden approach to its ecosystem services.

The move acknowledges that no single AI model provider can claim superiority across all use cases, and it positions Apple as a platform rather than a model builder in the AI stack. For Anthropic, Google, and OpenAI, the distribution implications are enormous: Apple Intelligence has been positioned as the primary on-device AI experience for over a billion iPhone users, and opening it to third-party models instantly creates a massive distribution channel for each company’s technology.

This development also signals growing competitive pressure on Apple to differentiate its AI offerings. With Google’s Gemini 3.5 family and OpenAI’s GPT-5.5 pushing the frontier of capabilities, Apple’s decision to embrace third-party models suggests the company recognizes that building the best AI internally is a different battle than building the best AI platform. For consumers, the ability to choose their preferred AI model within the Apple ecosystem is a significant quality-of-life improvement and may accelerate mainstream adoption of AI-assisted features on mobile devices.

3. Trump Abruptly Halts AI Executive Order Over China Competition Fears

President Trump abruptly halted the signing of a planned AI executive order, citing concerns that the regulations could weaken America’s competitive edge against China in the AI race. USA Today, AP News, and Crypto Briefing all confirmed the reversal, which came after Tech CEO David Sacks raised industry concerns about the proposed order’s regulatory burden. Politico reported that Trump yanked the AI order specifically after Sacks flagged the risks of overregulation.

The decision marks a dramatic about-face from the administration’s earlier signals that it was preparing to sign an AI cybersecurity directive. CNN and Bloomberg had reported just days earlier that the order was expected to be signed as soon as Thursday, characterizing it as a “cybersecurity directive” that would reshape how AI systems are secured across government and critical infrastructure. The reversal suggests that the tension between regulating AI for safety and maintaining American competitiveness in the global AI race remains unresolved at the highest levels of government.

Modern Diplomacy reported that Trump also delayed the AI executive order amid fears the US could lose ground to China in the artificial intelligence race. The Times of India noted that Trump’s China visit influenced his position on the Anthropic Mythos fears, which had been a key driver of the proposed regulatory framework. The cancellation leaves the AI industry in a state of regulatory uncertainty, with companies operating without clear federal guidance while state-level AI regulations continue to proliferate. Governing reported that states are increasingly split on AI policy, with Trump threatening lawsuits over state-level AI laws — further complicating the regulatory landscape.

4. Pentagon Drops Anthropic, Tests OpenAI and Google AI for Classified Military Work

The Pentagon has begun testing AI models from OpenAI and Google as replacements for Anthropic’s systems in classified military applications, marking a significant escalation in the ongoing dispute between the Trump administration and the AI safety-focused company. Bloomberg and the BBC reported that the Department of Defense is actively testing rival AI models after restricting Anthropic systems, following Trump’s order for US government agencies to stop using Anthropic technology.

DefenseScoop confirmed that the Pentagon has inked deals with eight AI companies for classified military work — notably excluding Anthropic from the list. The Guardian reported that the deals include major AI infrastructure providers, with TechCrunch specifically noting partnerships with NVIDIA, Microsoft, and AWS to deploy AI on classified networks. The Financial Times reported that the Pentagon is now an “AI-first” fighting force, signaling a fundamental transformation in how the US military approaches technology.

The exclusion of Anthropic is particularly significant given the company’s prominence in the AI safety space and its $40 billion investment commitment from Google, as reported by The New York Times. The Pentagon’s decision to test OpenAI and Google models instead reflects the administration’s preference for companies perceived as less focused on AI safety restrictions. CXO Digitalpulse confirmed that the Pentagon has begun testing OpenAI and Google AI models after restricting Anthropic systems, while Malwarebytes reported that OpenAI is effectively taking over the role Anthropic previously held in Pentagon AI operations. The shift has major implications for the AI safety debate: if the Pentagon — one of the most security-conscious organizations in the world — is willing to bypass Anthropic’s safety-focused approach, it could signal a broader industry shift away from safety-first AI development.

5. OpenAI Prepares Confidential IPO Filing — Valued at $850B+

OpenAI is preparing to confidentially file a draft of its IPO prospectus as early as this week, in what could become one of the largest public market debuts in history. The Wall Street Journal broke the exclusive report that OpenAI is preparing to file for an IPO “very soon,” while The New York Times confirmed the company “prepares to file for an I.P.O. in coming weeks.” Axios reported that OpenAI has prepared a confidential IPO filing, and Reuters cited sources saying the company is aiming for a speedy IPO timeline.

The company is now valued at more than $850 billion by private investors and is working with Goldman Sachs and Morgan Stanley to prepare the filing. CNBC confirmed the move on Wednesday, and an OpenAI representative issued a carefully worded statement: “As part of normal governance, we regularly evaluate a range of strategic options. Our focus remains on execution.” CFO Sarah Friar had previously told CNBC that it is “good hygiene” for a company of OpenAI’s size to “look and feel and act” like a public company, though she declined to comment on a specific timeline.

The timing is particularly notable because it coincides with SpaceX’s public IPO filing for its merged entity with xAI. CNBC reported that mega-IPOs from SpaceX and OpenAI could signal a market top, according to analysts, while CoinDesk noted that Tom Lee believes the SpaceX, OpenAI, and Anthropic IPOs could reshape markets. The dual filings represent an unprecedented moment in tech capital markets, offering investors two distinct visions of AI’s future: OpenAI focused on general-purpose AI models and applications, and SpaceX/xAI combining space technology with AI research infrastructure. TradingKey reported that OpenAI is facing off against SpaceX for the title of 2026’s most anticipated IPO.

6. Andrej Karpathy Joins Anthropic: OpenAI Co-Founder and “Vibe Coding” Pioneer Defects

Andrej Karpathy — one of the most prominent researchers in artificial intelligence, former OpenAI core team member, former Tesla AI director, and the widely credited inventor of “vibe coding” — has joined Anthropic. Axios confirmed the move, while TechCrunch reported that Karpathy will join Anthropic’s pre-training team. Fortune captured the significance of the move, describing Karpathy as the “inventor of ‘vibe coding’” who is defecting to Anthropic. CNBC reported that Anthropic has hired Karpathy, the OpenAI co-founder and former Tesla AI leader.

Karpathy announced the move in a post on X, saying he is “excited to get back into research and development” and calling the next few years at the frontier of large language models “especially formative.” He will be standing up his own pretraining team at Anthropic focused on using Claude to speed up pretraining research. Karpathy’s career trajectory maps the AI industry’s evolution: he was part of OpenAI’s core team in its early days, helped build Tesla’s Autopilot and Full Self-Driving technology, returned to OpenAI, and ultimately left in 2024 to work on AI in education through his startup Eureka Labs.

The significance of Karpathy choosing Anthropic over returning to OpenAI cannot be overstated. It represents a clear loss for his former employer and a major validation for Anthropic’s research direction. MarkTechPost noted that Karpathy’s role as a “Forward Deployed Engineer” at Anthropic represents the type of talent that OpenAI, Anthropic, and Google are all aggressively hiring in 2026. Karpathy recently said he was “blown away” by the progress of agentic AI for coding, after dismissing agentic capabilities just months before — a reversal that speaks to the accelerating pace of progress. His focus on using AI to improve AI pretraining represents a key thesis about exponential compounding in AI research that could reshape the competitive dynamics of frontier model development.

7. NVIDIA Invests Up to $3.2 Billion in Corning for AI Optical Fiber Infrastructure

NVIDIA has announced a massive partnership with Corning, investing up to $3.2 billion as part of a long-term deal to build three new optical fiber factories focused on AI infrastructure. CNBC reported the investment, which represents one of the largest single infrastructure commitments by NVIDIA to strengthen US manufacturing capacity for the AI supply chain. NVIDIA’s own newsroom confirmed the long-term partnership, emphasizing the strategic importance of domestic manufacturing for AI infrastructure components.

The partnership addresses a critical bottleneck in AI infrastructure: optical fiber connectivity. As AI data centers scale to hundreds of thousands of GPUs, the demand for high-bandwidth, low-latency optical connectivity has become one of the most constraining factors in AI deployment. Corning’s optical fiber technology is essential for the interconnect networks that link GPU clusters within and between data centers. The three new factories will significantly expand US domestic production capacity for this critical component.

Simply Wall St reported that Corning’s valuation has been reassessed following the NVIDIA partnership and the company’s 2026 growth guidance. 24/7 Wall St noted that NVIDIA might have revealed “the AI sleeper stock that investors missed,” suggesting the market is beginning to price in the significance of infrastructure partnerships beyond the obvious AI chip plays. The deal also has geopolitical implications: by strengthening US manufacturing for AI infrastructure components, NVIDIA and Corning are reducing dependence on foreign supply chains at a time when AI infrastructure has become a matter of national security. The Motley Fool noted that the Corning-NVIDIA partnership is reshaping the AI infrastructure investment landscape, with CoreWeave and Nebius facing increased competition for data center infrastructure deals.

📊 Trend Watch

🔭 What to Watch

• OpenAI’s IPO filing timeline — The confidential filing could arrive as soon as this week; the prospectus will reveal revenue, profitability timeline, and risk factors that will define how Wall Street values AI model companies.
• Apple’s iOS 27 model selection rollout — How users respond to choosing between Anthropic, Google, and OpenAI models within Apple Intelligence will be a critical indicator of brand loyalty vs. performance preference in consumer AI.
• Pentagon’s Anthropic replacement results — Whether OpenAI and Google AI models perform adequately in classified military applications will be a key data point in the safety vs. capability debate that has animated the Anthropic dispute.
• Trump’s next AI policy move — After halting the executive order, the administration’s next AI policy action will clarify whether the US is moving toward deregulation, targeted cybersecurity rules, or a hands-off approach.
• Karpathy’s pretraining team at Anthropic — Whether Karpathy’s thesis about AI-accelerated pretraining yields faster model improvement cycles could fundamentally reshape how frontier labs approach model development.