NYC AI Hearing: Top AI Bosses Face Council Subpoenas

Samantha: Welcome to The State of Tech, The European Edition, Monday October five, 2026. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: New York drags top AI bosses into a council hearing, Amazon tries to buy peace with billions for data centre towns, Hong Kong rides an AI fundraising wave, Google DeepMind watermarks lab-grown proteins, and to close out two things worth trying today, the free open-source audio tool Ocenaudio and the European route planner Naviki. Let's start in New York.

New York City Council hauls top AI bosses in to answer for the risks hitting ordinary citizens.

Samantha: The lead today sits in a council chamber in Manhattan, not Silicon Valley. Executives from OpenAI, Anthropic, Google, Meta and SpaceX's AI unit have been summoned before the New York City Council to testify on safeguards, risks to residents and possible local rules. The council warned it would issue subpoenas if anyone refused. That is a city, not a federal regulator, putting the biggest names in AI under oath about what their systems do to New Yorkers.

Bob: And that is the striking part. We are used to AI oversight happening in Washington or Brussels, not city hall. The council wants to understand how fast-moving systems affect housing, policing, hiring and public services inside one city, and what legislation is actually available at municipal level. It also signals impatience: lawmakers feel federal action is too slow, so they are stepping in where they can, even if their reach is limited to local procurement, consumer protection and public agencies.

Samantha: For the companies, the stakes are reputational as much as legal. A bad hearing produces clips that travel. And once a major city has a hearing record, it becomes a template. Other councils copy the questions, the witness list, the demands. That is how a local hearing turns into national pressure. It also forces the companies to put something on the record about safeguards they often prefer to describe only in blog posts.

Bob: The European lens is direct. European cities already have more regulatory room than American ones, through procurement rules, data protection authorities and the AI Act's layered obligations. If New York produces a workable script for grilling AI firms on concrete harms, expect Amsterdam, Berlin, Barcelona and Paris councils to borrow from it. The bottom line is that AI accountability is moving down from capitals to the municipal level, where the services citizens actually use are run.

Amazon pledges over a billion dollars to US towns hosting its AI data centres.

Samantha: Next up, Amazon is trying to defuse local anger around its data centre build-out. AWS chief Matt Garman unveiled a programme called Built Together, pledging more than one billion dollars over five years to US communities that host Amazon data centres. The money goes to free community college degrees, trades training and energy efficiency upgrades for local schools and homes. Amazon also promised its data centres will not raise local electricity bills and will publish annual figures on energy and water use.

Bob: The context is important. Across the United States, towns have pushed back hard against new data centres, worried about power draw, water consumption and strain on the grid. Several projects have been delayed or blocked outright. A billion dollars across five years and dozens of sites is not transformational per community, but the transparency pledge matters more. Publishing energy and water figures per site is something hyperscalers have long resisted, because the numbers invite exactly the scrutiny they want to avoid.

Samantha: The commitment on local electricity bills is the one to watch. Amazon is saying its data centres will not push up what residents pay. How that is measured and enforced is unclear, and utilities, not Amazon, set tariffs. If the pledge turns out to mean Amazon pays extra for grid upgrades rather than passing them to households, that is meaningful. If it is just a line in a press release, communities will notice quickly.

Bob: The European angle is sharp. Ireland, the Netherlands and parts of Germany and Spain already face moratoriums or restrictions on new data centres over grid and water concerns. European mayors will read this announcement and ask why they are not getting similar packages. Expect community investment and transparent reporting to become table stakes for any hyperscaler trying to open a new European site over the next two years.

Samantha: Quick interruption. If you listen to The State of Tech regularly, hit that like button and subscribe, that way you'll never miss an episode. Okay, moving on.

Hong Kong share sales hit a record forty-seven billion dollars as AI money floods in.

Bob: Hong Kong just posted its biggest quarter ever for share sales. Companies raised forty-seven and a half billion US dollars from July to September, pushing the year's total above ninety-two billion and closing in on the 2021 record. The driver is an AI-fuelled rush for capital, with Chinese companies lining up IPOs, placements and block trades to fund AI expansion. AI model maker Z dot AI raised a large round and returned to the market shortly after its initial offering.

Samantha: What makes this notable is where the money is landing. Hong Kong had a quiet few years for listings. The return of volume, specifically tied to AI, suggests investors are willing to back Chinese AI stories again despite the geopolitical overhang. The speed of the Z dot AI follow-on is telling. Companies are tapping markets fast because they expect capital needs to keep climbing, mainly for compute, data centres and talent, the same cost base pushing US AI valuations.

Bob: It also shows the global AI capital race is not a two-horse American affair. Chinese AI firms are financing themselves at scale on home-region markets, which reduces their dependence on Western funding and Western chips over time. That matters for the industrial balance. If Chinese AI companies can keep raising this kind of money, they can keep buying domestic alternatives to Nvidia hardware and sustain their own model development even under export controls.

Samantha: The European lens is uncomfortable. European exchanges have struggled to attract large AI listings, with most ambitious companies looking to New York or now Hong Kong. If capital for European AI keeps flowing out, the sector risks becoming a research base that lists elsewhere. Industry groups argue Europe needs deeper pools of growth capital and simpler listing rules to compete, otherwise the next wave of AI champions raises its money somewhere else and reports to shareholders somewhere else too.

Google DeepMind watermarks AI-designed proteins without breaking their function.

Bob: Our fourth story moves into the lab. Google DeepMind says it has created watermarked AI-designed proteins that still work in biological tests. The watermark is an invisible signature embedded in the protein's structure, letting researchers verify that a given molecule was generated by AI, without the watermark compromising its biological activity. This matters because AI protein design is moving fast into drug discovery, enzymes and new materials, and provenance has become a serious problem.

Samantha: Until now, once an AI designed a protein and someone synthesised it, there was no reliable way to tell AI-generated structures from naturally derived or human-designed ones. That created real headaches: for intellectual property, for safety screening, and for biosecurity, where you want to know whether a dangerous sequence came out of a model. A functional watermark gives regulators and labs a hook. It is the biological equivalent of signing an image or a document.

Bob: The caveat is that this is a research result, not a deployed standard. For a watermark to matter, it has to be adopted widely, including by companies that may prefer their designs stay unmarked. DeepMind publishing the method is a push toward making it a norm. The functionality test is also crucial. Earlier watermarking attempts in text and images often degrade output quality, so showing proteins still work in the lab after watermarking is the actual breakthrough here.

Samantha: The European lens is real. European biotech hubs in Basel, Cambridge, Munich and Leuven are heavy users of AI protein design, and European regulators under the AI Act and medicines rules are already asking how to verify AI-generated biological content. A workable watermark gives them something concrete to point at. If DeepMind's approach holds up to independent testing, expect it to show up in European safety and provenance guidance within the next year or two.

The free open-source audio editor Ocenaudio turns any laptop into a quick, clean sound workbench.

Bob: To close out, two things you can actually use today. The first is Ocenaudio, a free audio editor for Windows, Mac and Linux. It handles the jobs most people actually need: trimming a voice memo, cleaning up background noise on a Zoom recording, cutting a music file to length, applying real-time effects while you preview. Users praise it for opening huge audio files quickly, where heavier editors stall or crash on long recordings.

Samantha: What makes it a useful pick is the gap it fills. Full audio suites are powerful but intimidating, with hundreds of menus and a learning curve measured in weekends. Ocenaudio sits in the middle. It is simpler than the professional tools, more capable than the basic recorder built into your operating system. For anyone who occasionally needs to edit a podcast clip, a voice note or a lecture recording, it is the kind of tool you install once and keep around for years.

Bob: Practical notes. It is free, with no account required and no subscription. It runs on Windows, macOS and Linux, which is unusual for audio software. It supports common formats like MP3, WAV, FLAC and OGG, and lets you preview effects before applying them, which saves time on trial and error. Installation takes a minute. Reviewers describe it as the tool you reach for when you do not want to launch a full studio just to cut thirty seconds off a file.

Samantha: The European angle is practical. Because Ocenaudio runs entirely on your own machine, your audio never leaves your laptop. For journalists handling source recordings, teachers editing lesson material or anyone cleaning up personal voice memos, that keeps sensitive audio off third-party servers by default. No cloud upload, no processing on someone else's infrastructure. It is a small privacy benefit that comes almost by accident, simply because the tool is built as a local app rather than a web service.

Naviki is the free cycling route planner from Germany that steers you onto quiet, bike-friendly roads your phone's default map will not.

Bob: Second pick is Naviki, a German route planning app built specifically for cyclists. You enter where you are starting and where you want to go, and it plans a bike route that prefers cycle paths and quieter roads over the fastest way by car. It is a different kind of navigation tool from the mainstream apps most of us have on our phones.

Samantha: The reason to try it is that general navigation apps are built for cars. Ask Google Maps for a cycling route and you often get something that technically works but puts you on a busy road for half of it. Naviki was designed for people on bikes. Users value that it favours dedicated cycle infrastructure, and that its routes are meant to be pleasant as well as direct.

Bob: Practicalities. Naviki is available as an app for iOS and Android and also works in your web browser, so you can plan a trip on a big screen at home and then take it with you on your phone. You can create a free account to save your routes and sync them across devices. Voice guidance is there for turn-by-turn directions on a bike, where glancing at a screen is not safe. The community element helps too, since other cyclists share routes you can simply follow.

Samantha: The European angle is built in. Naviki comes from Germany, a country with a strong cycling culture, and it draws on OpenStreetMap data, which is detailed across much of the continent. Its network follows the same idea you will find in many European cities, where bike routes are treated as real infrastructure rather than an afterthought. If you have been meaning to explore your region by bike this month, it is a low-friction way to find routes that are not the obvious ones.

Samantha: Today we covered: New York council grills top AI executives, Amazon pledges over a billion for data centre communities, Hong Kong's record AI fundraising quarter, DeepMind's watermarked proteins, and two things to try yourself: the free audio editor Ocenaudio and the European route planner Naviki.

Bob: Want to know more or react? Visit stateoftech.eu or email us at info@doorzetters.net.

Bob: State of Tech, the tech world in 15 minutes.