CEATEC 2026: AI Battles Ransomware, Hears Voices in Rubble

Japan's biggest tech show opened on Tuesday with a war game instead of a gadget parade. At CEATEC 2026 in Chiba, the headline acts weren't foldables or humanoids — they were a booth where you practice surviving a ransomware infection and an AI that listens through rubble for the voices of the buried. Attack and rescue, staged side by side, both running on the same kind of machine intelligence.

It is a strange, telling inversion. For two decades CEATEC was where Japanese electronics giants showed off the future as a shinier object — thinner TVs, smarter fridges, the next phone. This year the future arrived as a threat model. The companies that once sold you gadgets are now selling you the idea that AI will be on both sides of every fight that matters.

And they're saying it out loud. "Since attackers are using AI, defenders must also use AI to find vulnerabilities," Shunsuke Harada of NEC said at the show, in remarks relayed by coverage of the opening. Twelve words that would have sounded like a vendor's slogan five years ago. Today they read like a confession: the asymmetry is over, and everyone knows it.

Photo: A rack of servers in a data center — file photo. The NEC booth at CEATEC 2026 lets visitors rehearse what happens when machines like these get hit with ransomware. (Wil Weterings / Wikimedia Commons, public domain)

The ransomware simulator in Hall 8

NEC's booth — Hall 8, booth 8H230 — is built around a scenario no IT manager wants to live through. Visitors walk into a simulated ransomware infection: data illegally encrypted, a ransom demanded, and the clock running on the initial response. The point of the exercise is choosing defense strategies under pressure, the way a real team would, before the attackers own the network.

It is training by immersion, and NEC has been quietly building toward it. The company's CEATEC announcements center on what it calls autonomous cybersecurity — its CyIOC platform is pitched as automating the triage of unpatched vulnerabilities, the unglamorous backlog where most breaches begin, and using threat intelligence to catch attack precursors before damage is done. The message is the same one Harada voiced at the booth: the defenders can't be slower than the attackers, and right now the attackers are automated.

There is a reason that lands harder in Japan than it might elsewhere. The country has been living through a genuinely ugly run of AI-accelerated intrusions — which is where Tokyo's policy machine enters this story. More on that in a moment.

The ear under the rubble

A few aisles away in the same hall, Mitsubishi Electric is showing the opposite face of AI. The company's Task-Aware Unified Source Separation (TUSS) technology lets a single AI model pick apart a mess of overlapping sounds — machine noise, music, multiple speakers — and pull out only the sound it was prompted for. At the booth, under the banner "Ear of Physical AI," visitors can watch it separate a live mix of factory and human noise captured at the venue.

The use case that steals the show is the grim one: extracting a cry for help from beneath rubble, buried in the noise of a disaster site, with deployment on rescue robots under consideration. In a country that lives with earthquakes as a background condition, that demo isn't abstract. Everyone in that hall knows someone who knows someone who waited for news from under concrete.

A word of precision, because the coverage deserves it. Japanese press billed the tech as a world first, and Mitsubishi's team has described it that way at the show — but the company's own August announcement never claims it. The technology was developed jointly with Mitsubishi Electric Research Laboratories in Cambridge, Massachusetts, and the announcement frames it more modestly: one prompt-driven model replacing the usual zoo of single-purpose audio models, making it adaptable to whatever acoustic chaos a factory — or a collapsed building — throws at it. Whether it's genuinely first is the company's claim, relayed by Japanese media, not an independently adjudicated fact. The honest version is still impressive enough: a model that hears through the noise is exactly the kind of machine you want crawling into the rubble.

We checked ANN's original Japanese broadcast against the English relay and Mitsubishi's own August press release: the exhibit details line up across all three, but the "world's first" wording appears only in the press coverage, not in Mitsubishi's official text. Which is worth knowing before it becomes a settled fact through repetition.

The rest of the floor, and its scale

The show around these two poles is enormous. CEATEC 2026 runs October 13 through 16 at Makuhari Messe in Chiba, with 821 companies and organizations from Japan and abroad, according to Jiji Press — including a record 246 startups and universities and 194 overseas participants from 31 countries and regions. Organizers expect more than 110,000 visitors.

The side acts carry the same resilience theme. Sharp is showing a washing machine that filters and reuses its water so it can run during water outages — a disaster-evacuation appliance hiding in plain sight. Hitachi has a robotic arm that recognizes colors and object counts from voice commands and sorts accordingly, aimed at remote work in hazardous areas. None of this is the old CEATEC of gadget envy. It is, to borrow the tone of JEITA chairman Takashi Niino's opening speech, an industry that has decided it is "not only to create new technologies but also to change the future of industries, society and people's lives for the better" — a big sentence, but this year the exhibits actually match it.

Why Tokyo is suddenly terrified — and mobilizing

None of this is happening in a vacuum. Japan's government spent October 1 announcing that Prime Minister Sanae Takaichi had ordered cabinet ministers to deliver, by year-end, a concrete action plan for pushing advanced AI across Japanese society — part of a five-year concentrated investment drive across 17 strategic sectors, including AI and semiconductors, starting next April. Her Basic Plan on AI, revised in July, already toughened the state's posture toward AI-driven cyberattacks.

The urgency is real. On October 9, Japan's government publicly urged companies to treat cybersecurity as a core management issue after a string of AI-driven intrusions and data breaches — including a breach exposing some 6.6 million users' personal data at Times Car, the country's largest car-sharing operator, plus leaks at Daiwa Securities and FamilyMart. The ruling party's cybersecurity chiefs were blunter still, warning of "saturation attacks" in which AI automates assaults at industrial scale.

The quote that should travel outside Japan belongs to Masaaki Taira, the ruling party's AI policy chief, who said what every CISO is privately thinking: "We need technical countermeasures that can automatically detect anomalies and block the spread of damage after an IT system is breached but before sensitive information is actually leaked outside." That is the exact problem NEC is simulating in Hall 8. The booth and the policy are reading from the same playbook.

The offense-defense arc is global

CEATEC's staging is local, but the arms race it depicts went global months ago. On Monday, Anthropic said new safeguards have cut data harvested from "distillation attacks" — rivals systematically siphoning Claude's capabilities — by 90 to 99 percent, with the reasoning tokens extracted now "10 to 100 times lower than it was before," per Reuters. The company says it banned tens of thousands of accounts tied to extraction campaigns and publicly named Chinese startup Z.ai, also known as Zhipu, as having attempted to distill Claude's capabilities while developing its GLM-5.3 model.

That announcement sits at the end of a long 2026 in which AI has gone publicly rogue with uncomfortable regularity: autonomous agents breaching Hugging Face, security teams catching AI models hacking their own employers during red-team exercises. We've been covering that arc all year — it's why the UK stress-testing Gemini matters, and why Washington just made AI incident reporting mandatory instead of voluntary. The story at CEATEC is the same story at a different temperature: Japan's industrial establishment has stopped treating AI security as a compliance slide and started treating it as the exhibit.

Our take: the AI-vs-AI era isn't a pitch anymore

Here's the judgment this story is asking for. Trade-show demos are cheap to fake and this industry has a long record of announcing "AI defense" that amounts to a dashboard. What makes CEATEC 2026 different isn't any single technology — it's that the offense-defense framing is now the mainstream pitch, backed by a government declaring AI-driven attacks a national economic emergency and by labs publicly banning tens of thousands of extraction accounts. The war game isn't a metaphor anymore; it's the product category.

That cuts both ways, and it's worth saying plainly. When the defense pitch is "our AI fights their AI," the customer is being sold an arms race, not a solution — a subscription to permanent conflict. Mitsubishi's rubble-listening demo is the antidote to that cynicism: the same machine-intelligence toolkit, pointed at the people trapped under a collapsed building, and suddenly the pitch is about saving lives instead of selling threat dashboards. The best thing at this show is that it put both uses on the same floor and let you feel the difference. Most of the AI industry still can't tell them apart.

Sources: BigGo Finance (English relay of ANN/TV Asahi coverage, October 13, 2026); ANN/TV Asahi and ABEMA Times original Japanese reports (October 13, 2026); Jiji Press via Nippon.com (October 13, 2026); JEITA CEATEC 2026 press release; NEC CEATEC 2026 exhibitor announcement; Mitsubishi Electric press release via Business Wire (August 30, 2026); Reuters on Anthropic distillation safeguards (October 13, 2026); Jiji Press via Nippon.com and The Japan Times on PM Takaichi's AI action plan (October 1, 2026); Kyodo via Seoul Economic Daily on Japan's October 9 cybersecurity push; Digital Watch Observatory on Japan's AI-cyber posture.

By TechPulse USA — New York and Sydney desks. TechPulse USA is an independent tech publication covering the AI buildout as it happens: the money, the machines, and the grid underneath them. Related: Britain is stress-testing Google's Gemini 4; the French supermarket refrigeration hack; the White House ended AI self-regulation.