AI co-scientists speed drug leads & Musk vs OpenAI trial fallout - Tech News (May 20, 2026)

An AI system just helped researchers surface promising drug candidates in hours—work that usually drags on for weeks—and it’s forcing a rethink of what “research assistance” even means. Welcome to The Automated Daily, tech news edition. The podcast created by generative AI. I’m TrendTeller, and today is May 20th, 2026. Let’s get into what happened, and why it matters.

AI co-scientists speed drug leads

First up in science: two separate reports in Nature are putting a spotlight on what’s being called AI “co-scientists.” The big idea is not a single chatbot, but a team of specialized AI agents that can scan literature, propose hypotheses, sketch experiments, and help interpret results—while humans still decide what’s worth testing and do the lab work. In one test, Google DeepMind’s system was used for drug repurposing work related to acute myeloid leukemia, generating candidate options quickly. Researchers picked a handful to try, and a few showed encouraging early signals in cultured cells. Another system from the nonprofit FutureHouse, nicknamed Robin, worked through dry age-related macular degeneration and flagged an existing glaucoma drug as a plausible candidate, along with suggestions for follow-up assays. The significance is speed: parts of early discovery that used to take weeks can be compressed to hours or days. The caution is just as important—cell-culture hits fail all the time once you move into harder tests—so this is acceleration, not a magic “drug button.”

Musk vs OpenAI trial fallout

Staying with research, Google researchers also described a separate Nature result: an AI system that can generate and refine scientific software—sometimes beating human-written code on tasks where you can score outcomes numerically. Think forecasting hospitalizations, modeling neural activity, or improving methods for analyzing biological datasets. What’s interesting here isn’t that AI writes code—that’s old news—but that the system is judged by results, not style. It tries many variants, keeps what works, and iterates quickly. If it holds up outside demos, it could shift a lot of scientific work from wrestling with pipelines to spending more time on the actual questions.

Google I/O: Gemini everywhere

Now to the AI industry’s ongoing identity crisis, with a courtroom twist. The California trial between Elon Musk and OpenAI CEO Sam Altman has ended in a verdict that largely favored OpenAI, with Musk framed as losing on a technicality. Beyond the personal drama, the subtext is bigger: the legal system effectively treated aggressive competition and profit-seeking in AI as normal business behavior—not as some special betrayal of early “for humanity” rhetoric. That outcome also reduces immediate pressure on OpenAI’s finances, clearing a path toward more fundraising and, potentially, a future IPO. But the trial didn’t settle the questions many people actually care about—like governance, accountability, and who gets to steer these systems. If anything, the spectacle may have further dented public trust by making the industry look like a power struggle among a small circle of executives.

Watermarking expands across AI media

On the money-and-momentum front, Polymarket is moving deeper into private-market speculation. It’s launching prediction contracts tied to private-company milestones—things like valuation thresholds, IPO timing, and secondary market activity—for high-profile AI firms. No, it’s not equity, and buyers don’t get shareholder rights. But it does create a public signal in a world where private pricing is often opaque. One key detail: Nasdaq Private Market will be the exclusive source to resolve outcomes, and it plans to publish relevant valuation data publicly for free as part of this rollout. If that sticks, it could nudge private markets toward a little more transparency—though it also adds one more way for hype to move fast.

GitHub breach via poisoned extension

Let’s talk Google I/O 2026, because the theme was clear: Gemini is becoming the default layer across Google. Google announced Gemini-driven upgrades spanning Search, the Gemini app, Workspace, YouTube, and shopping—plus a new Android XR category for what it’s calling intelligent eyewear. Search’s AI Mode is moving to a new Gemini Flash model and is leaning harder into ongoing, conversational tasks—like creating persistent trackers and letting background agents monitor things you care about. The Gemini app itself is getting a redesign aimed at more structured, visual answers, and Google is pushing “agent” features that can take actions across apps—starting with Workspace—along with a daily briefing that pulls priorities from Gmail, Calendar, and Tasks. Meanwhile YouTube is getting better AI-based video finding, and there’s a bigger push to help creators generate Shorts using Gemini-powered tools. It’s a lot, but the story is simple: Google wants AI to feel less like a destination and more like an operating system for your day.

China accelerates brain-computer interfaces

One quieter but important thread from Google: watermarking and provenance. Google says its SynthID watermarking system has now labeled an enormous amount of AI-generated media, and—more importantly—it’s expanding beyond Google’s own models. Partners announced include Nvidia for some of its model tooling, plus OpenAI for its image generation, and others in audio and consumer platforms. The takeaway: watermarking only becomes truly useful when it’s shared infrastructure. It won’t catch everything—open models and unmarked content will remain a gap—but wider adoption could make it easier for platforms and users to sanity-check what they’re seeing, especially as synthetic video keeps improving.

Polymarket bets on AI startups

Now, a security story with a very modern lesson: GitHub says it’s investigating unauthorized access to internal repositories after a compromise of an employee device. The intrusion was traced to a poisoned version of a VS Code extension. GitHub says its current assessment is that the attacker accessed GitHub-internal repos, and it has no evidence so far of customer data exposure outside those internal repositories. The company rotated critical secrets and is monitoring for follow-on activity. The bigger point is supply-chain risk at the developer-tooling layer. Extensions and plugins are incredibly powerful—and increasingly, they’re a soft target.

Apple Siri adds auto-delete privacy

On the future-of-humans front—literally—Chinese startups are accelerating AI-powered brain–computer interfaces and moving from small trials toward products intended for broader public use. Some early results include cursor control for people with paralysis and reported progress decoding Mandarin speech signals in a patient scenario. China’s government is also backing the space with explicit targets for breakthroughs by 2027 and ambitions to grow world-class BCI companies by the end of the decade. This is promising for accessibility and treatment, but it intensifies a hard question: what does informed consent mean when your neural signals become training data, and how do you prevent brain data from becoming the most intimate form of surveillance?

Epic vs Apple fight expands

In platform power and regulation, Epic says Fortnite has returned to the App Store in every country except Australia, casting it as the next phase of its long-running fight with Apple. Epic’s argument is that global regulators are watching Apple’s commission rules closely, and a new round of court scrutiny in the U.S. could reveal how Apple justifies its fees—especially when purchases happen through external links. Australia is the holdout because Epic says Apple is still enforcing terms an Australian court found unlawful. This isn’t only about one game; it’s part of the broader contest over who controls distribution, payments, and policy on the phone in your pocket.

Science and journalism vanish online

Two quick notes to close. First, Apple is expected to add new Siri privacy controls in iOS 27, including options to automatically delete Siri conversation history after a set period. It’s a simple feature with a big signal: assistants are becoming the place people share sensitive details, and “keep everything forever” is starting to look like the wrong default. And finally, a reminder that the internet forgets—sometimes on purpose. Nate Silver says Disney/ABC has effectively wiped much of FiveThirtyEight’s Disney-era archive from the open web, with old links now redirecting to ABC News’ homepage. It’s an extreme case of link rot, and it raises an uncomfortable question: if major institutions can erase years of reporting and interactive work overnight, what does that do to our shared record?

That’s the Automated Daily, tech news edition for May 20th, 2026. If one theme tied today together, it’s this: AI is accelerating science and products at the same time that trust, governance, and preservation are struggling to keep pace. Thanks for listening. I’m TrendTeller—see you tomorrow.

AI co-scientists speed drug leads & Musk vs OpenAI trial fallout - Tech News (May 20, 2026)

Our Sponsors

Today's Tech News Topics