The past three days have offered a vivid snapshot of where the AI industry stands in late 2026: enormously powerful, deeply entangled with politics and markets, and still capable of surprising even its own creators.

Wall Street's AI Trade Unravels

The biggest story isn't a product launch — it's a market reckoning. Following the U.S. midterm elections, in which Democrats won control of both chambers of Congress, investors dumped AI-linked stocks on fears that the incoming Congress and newly elected state officials will slow the data center buildout that has powered the sector's valuations. Between January 2024 and May 2026, at least 46 AI data center projects across 20 U.S. states, representing more than $170 billion in announced investment, were blocked, withdrawn or stalled following community opposition. The electoral results amplified that backlash: according to Data Centers on the Ballot, 39 ballot measures to restrict AI were successful, while only one — in North Dakota — voted against restrictions.

The damage has been staggering. AI stocks have shed approximately $13.2 trillion in market value since their 2026 highs, with memory stocks down an average of 60% from their 2026 highs and AI power stocks falling an average of 53%. Analysts at Goldman Sachs reportedly warned that Democratic wins at the state and local level are likely to lead to short-term data center moratoria, potentially affecting about 45% of planned US data center capacity expansion. It's a reminder that the AI boom's physical infrastructure — power-hungry data centers — has become a genuine political flashpoint, not just a technical one.

DeepMind Turns AI Agents on Cryptography

Amid the market chaos, Google DeepMind delivered a research result with long-tail implications for security teams. Google DeepMind has published new research showing that AI agents, working under human supervision, can dramatically accelerate cryptanalysis, uncovering two significant weaknesses in legacy 1024-bit RSA cryptosystems. One breakthrough targets digital signature forgery; the other speeds up batch factoring of many keys at once, with DeepMind's agents finding a fan factory method that factors each extra 1024-bit key at about 1/58 of a GNFS run.

Crucially, modern RSA deployments using 2048-bit or larger keys remain secure against this specific threat, as the research targets the older, shorter keys, not the current standard. DeepMind VP Pushmeet Kohli framed it as a methodology story as much as a security one, noting "these results show why we must not assume the mathematical 'hardness' of cryptographic schemes which we have relied upon in the past." It's a small but telling data point in the broader "AI for science" push — agents compressing years of specialist research into weeks.

Cheaper, Longer AI Video

On the product side, Google quietly made one of its more practical upgrades of the year. Google rolled out Gemini Omni 1.2 Flash, making it generally available across the Gemini API and Gemini for business, with native 15-second video creation, multi-turn conversational editing, and a 30% price cut from Gemini Omni 1.1 to $0.07 per second for 720p video. It's a pure speed-and-cost play rather than a capability ceiling push, but for startups building video tools, a 30% price cut compounds fast at scale.

Taken together, these stories capture AI's current moment: breathtaking capability gains running headlong into financial and political gravity.

Sources

AI StocksArtificial IntelligenceCybersecurityDaily AI UpdateGeminiGoogle DeepMind