Denmark’s ID Registry Breached as Trump Unveils Super Intelligence Force

AI & Machine Learning

Independent researchers monitoring internet traffic have spotted a fleet of AI agents running on Tencent infrastructure and repeatedly querying Alibaba’s Amap mapping service for directions to parks, zoos and hospitals, effectively sidestepping Amap’s API restrictions. The activity surfaced through URLquery, a domain-scanning service whose logs reveal the sites agents touch indirectly. The researchers deliberately call it an “agent fleet” rather than a swarm, describing many parallel agents doing the same task with no sign of coordination between them. The behaviour does not appear malicious, but it shows how pervasive and lightly concealed autonomous agent traffic has become, and why researchers are now watching for it after a string of agent-related incidents this autumn. Source: TechCrunch Verified: True

Anthropic has begun prompting Claude users of its voice features to share audio recordings and voice-chat data so it can train its models to better understand and respond to speech. The program is strictly opt-in: users see an “Allow” or “Not now” prompt, the setting is off by default, and a dedicated toggle sits in Claude’s privacy settings. Voice training consent is separate from the existing chat and code-session training options, giving users granular control, and shared voice data can be deleted at any time. The move signals how labs are hunting for high-quality spoken-language data as voice becomes a primary interface for assistants. Source: BleepingComputer Verified: True

Startup Lola Vision Systems is building a compiler toolchain that translates trained AI models into instructions specific edge chips can execute, automating a porting job its founder says takes roughly 200 hours by hand. Founder Tayo Adesanya, who previously helped manufacturers select AI processors, is also developing proprietary silicon but plans to license the software for existing hardware first to bring in revenue sooner. The company has raised just over $1 million, has one signed customer and about a dozen letters of intent, mostly from aerospace and other mission-critical sectors. Its pitch targets a real bottleneck in edge AI, where accuracy and reliability on constrained chips can decide whether a product clears regulatory review. Source: TechCrunch Verified: True

Consumer Hardware

OpenAI is adding visual display ads that appear alongside image-generation results in ChatGPT for users on its free and low-cost tiers, starting in the United States with a first advertiser test group later in October. The company says the ads are clearly labelled and will not influence ChatGPT’s answers, and it has signed up measurement and brand-safety partners including AppsFlyer, Adjust, Triple Whale, DoubleVerify and Integral Ad Science. OpenAI is also running geo-based ad experiments with Haus, Measured and WorkMagic, and plans an eventual global rollout across its roughly 1.2 billion weekly users. The expansion builds on text ads introduced earlier this year and an August launch in India, deepening ChatGPT’s turn into an ad-supported consumer platform. Source: TechCrunch Verified: True

Canonical has shipped the beta of Ubuntu 26.10 “Stonking Stingray” ahead of a final release scheduled for October 15. The interim release continues Canonical’s “oxidization” drive to replace C-based tools with Rust equivalents, with OpenPGP support now handled by the Rust-based Sequoia-PGP library. It also introduces Myna, an optional LLM-powered speech-to-text tool that lets users pick from several recognition models, and ships the GNOME 51 “A Coruña” desktop. Hands-on testing found the beta ran smoothly on virtual machines and bare metal, apart from cosmetic theme errors, with an installer that now fetches updates during setup. Source: The Register Verified: True

Cybersecurity

Attackers stole the records of about 8 million people from Denmark’s Central Person Register, the national database holding roughly 11 million entries, in what is considered the country’s largest breach on record. The stolen data includes names, addresses, CPR social security numbers and government-issued identity numbers. Rather than breaking in directly, the attackers abused a Danish company’s lawful access credentials to search the register; the intrusion took place in September and was discovered on October 2. Minister Christina Egelund called it a “serious incident,” and the government has declined to name the attackers, drawing comparisons with past national-ID leaks in Turkey and India. Source: TechCrunch Verified: True

South Korea’s Financial Services Commission convened an emergency meeting after confirming cyberattacks on several of the country’s largest banks, including Shinhan Bank, where data on 25,000 customers leaked, and KB Kookmin Bank, where credit card details of 119,000 customers were exposed. Hana Bank reported a narrower breach of a sales-support system. Regulators launched on-site investigations and ordered financial firms to audit internet-facing systems, tighten authentication and share threat intelligence. Korean media reported an HTML page title referencing ARTEX AI, an open-source automated penetration-testing framework, fuelling suspicions that AI attack tooling was involved, though authorities have not confirmed the link. Source: BleepingComputer Verified: True

Dell has warned PowerEdge server customers to patch CVE-2026-86360, a critical path-traversal flaw in its Dell System Update command-line deployment tool that lets an unauthenticated remote attacker execute code with root privileges. Dell says successful exploitation means complete compromise of the application and underlying operating system, and the fix is DSU version 2.3.0.0 or later. The same release closes four further high-severity bugs enabling remote code execution and privilege escalation. No exploitation has been reported yet, but state-backed groups have weaponised Dell flaws before, and CISA has long branded path traversal bugs as unforgivable. Source: BleepingComputer Verified: True

Google has frozen its Open Source Software Vulnerability Rewards Program, citing “a significant rise in automated submissions, the vast majority of which are not valid.” The pause took effect on October 1, and Google says it will provide an update in the first quarter of 2027. Engineers and open-source maintainers had been swamped by reports containing AI hallucinations, mirroring warnings from security experts that machine-generated slop is undermining bug bounties. The decision removes a key incentive for finding flaws in widely used open-source code at the same time attackers are increasingly using AI to hunt for real ones. Source: TechCrunch Verified: True

Enterprise Infrastructure

BT Group has bought TalkTalk Telecommunications and PlatformX Communications out of administration on a debt-free basis, keeping broadband running for 2.5 million consumer and wholesale customers at an estimated £400 million cash impact this financial year. TalkTalk generated around £1.2 billion in revenue over the past 12 months but was loss-making, and a long sale process failed to find another buyer. Digital Secretary Lisa Nandy invoked public interest powers, calling phone and broadband vital national infrastructure, and the Competition and Markets Authority must report on competition concerns by October 19. Virgin Media blasted the deal as “a stitch up masked as a rescue,” arguing BT should face the same scrutiny applied to rival consolidation days earlier. Source: The Register Verified: True

Windows 11 has climbed to 71.5 percent of Windows desktops versus 27.77 percent for Windows 10, according to Statcounter, as the first year of Microsoft’s paid Extended Security Updates program for businesses comes to an end this month. Commercial ESU pricing doubles at each renewal, from $61 per device in year one to $122 in year two and $244 in year three, the maximum term. Consumers, Windows 365 users and residents of the European Economic Area get updates free or via other routes, so the bill lands squarely on enterprises. Analysts at Lansweeper say the steep part of the migration curve is over, with remaining Windows 10 fleets largely machines where paying ESU is still cheaper than new hardware. Source: The Register Verified: True

Policy & Regulation

President Trump has formally unveiled his “Super Intelligence Force,” a federal task force to coordinate US government AI development and safety work under the administration’s new “super intelligence” branding. Director of National Intelligence Jay Clayton chairs the group, with FTC Chair Andrew Ferguson, Undersecretary of War for Research and Engineering Emil Michael and OPM Director Scott Kupor as vice chairs. The force must deliver a report within 120 days on AI opportunities and risks while “preventing overregulation and regulatory capture that would stifle innovation and competition.” Clayton framed falling behind China as a bigger danger than the technology itself, signalling that the body will lean toward acceleration rather than restraint. Source: TechCrunch Verified: True