AI spent the week animating pictures, eyeing your inbox, and sending poetry from orbit. Back on Earth, Apple was trimming AI allowances, voters were putting limits on data centers, and attackers were finding fresh ways into phones and cloud accounts. The future keeps arriving with impressive demos and a growing list of maintenance requirements.
Top news
Generative AI moves into images, apps, and browsers
ChatGPT Images 2.6 adds the ability to create looping GIFs from text prompts or uploaded photos. Users can adjust playback speed and preserve reusable references, while image generation is reportedly 50% faster. The GIF feature is rolling out gradually, however, and early testing identified continuity flaws.
Apple has released a Siri AI developer preview built around App Intents and new app skills. The framework lets developers create custom, multistep requests, but integrations are currently restricted to developer builds and internal TestFlight testing. App Store submissions are expected to open in spring 2027.
Chrome is gaining Gmail-powered Gemini autofill that can retrieve flight references, tracking numbers, and order IDs from a user’s inbox when completing online forms. Google is gradually introducing the feature on desktop, Android, and iOS in the United States.
AI policies reach schools and research labs
Beginning in early 2027, managed K-12 accounts must accept Google’s integrated AI features to retain access to Search. School administrators will be able to separate Search from Gemini chatbots, but they will not be able to provide traditional Google Search without generative AI functionality.
Anthropic is warning that future models could significantly accelerate AI research and potentially move faster than existing safety checks. AI was responsible for 26% of Anthropic’s model research and development work in August under human supervision, while the share of fully autonomous work remained at 0%.
AI hardware expands from wearables to orbit
Samsung’s display-free Android XR glasses could launch in the United States and South Korea in mid-December, although the company has not confirmed the timing or price. Equipped with cameras, microphones, and speakers, the frames are expected to offer Gemini-powered translations, directions, and other forms of assistance.
Apple is reportedly considering a $1,000 tabletop robot featuring a 9-inch smart display mounted on a robotic arm. A launch could come in 2027 or 2028. Its software team is reportedly pushing to narrow the device’s scope to FaceTime and home automation.
Google’s Project Suncatcher satellite generated a haiku from orbit using an onboard Gemma 4 model, marking the first response from the company’s space-based AI experiment. The project must still overcome significant power, radiation, thermal, networking, and cost challenges.
Plant-based products reach retail shelves
Steakholder Foods, a maker of 3D food printers, has launched plant-based meat and seafood products in stores across the northeastern United States. It remains unclear, however, which of the retail products were actually produced through 3D printing.
Insider intel
Enterprises reassess the cost and value of AI
Apple reportedly cut engineers’ daily generative AI allowance from $300 to $200 and could impose further reductions. Amid rising AI and infrastructure expenses, the company is also said to be directing some teams toward less expensive models or on-device processing.
Amazon is warning developers against maximizing token consumption simply to improve internal AI usage metrics. Its guidance emphasizes measurable results rather than raw consumption and acknowledges that some engineering tasks may not benefit from AI at all.
Together, the moves illustrate a growing focus on AI efficiency. Even as companies broaden access to generative systems, internal leaders are paying closer attention to infrastructure costs and whether increased model use produces meaningful business outcomes.
Security alerts
Mobile devices face web and messaging attacks
A campaign linked by Sekoia to Russia’s APT29 compromised an analytics script used by thousands of legitimate websites to target unpatched iPhones and Android devices. Users should install all available operating-system and browser updates to reduce their exposure.
The Android vulnerability CVE-2026-58882, described as an “SMS of Death,” allows invisible malicious text messages to crash cellular calling services. An attack could drop emergency calls or send affected devices into reboot loops. The fix covers Android 14–17, but availability varies by manufacturer. Install the latest available security update; Android 12–13 users should check with their device maker about patch support, as unsupported devices may remain exposed.
Credentials and AI access become high-value targets
Password manager mSecure detected suspicious activity, and a customer-posted support reply says an attacker stole an encrypted database containing data from current and former mSecure Cloud sync users. Developer mSeven says there is no evidence the confidential contents were exposed. Customers are being advised to change their master and individual account passwords, replace stored passkeys, and re-enroll two-factor authentication secrets.
Attackers are also using stolen cloud credentials to activate enterprise AI services and resell access through illicit “LLM Mixers.” Researchers observed more than $100,000 in unauthorized usage in a single day, while some victim organizations accumulated charges as high as $1 million.
Binance has linked a $6.3 billion cryptocurrency theft to a fraudulent job interview assignment that delivered malware. According to the company, attackers compromised engineers, stole deployment credentials, and reached wallet-signing infrastructure. The FBI attributes the operation to TraderTraitor, a group linked to North Korea.
Malicious software exploits trusted platforms and products
Researchers identified 319 Chrome extensions with 13.4 million reported installations that redirected searches through third-party servers to generate referral revenue. Users should remove affected add-ons, restore their preferred default search engine, and run an antivirus scan.
Fake ChatGPT and Claude applications distributed through GitHub are installing malware on Windows PCs. The malicious downloads combine legitimate, signed installers with a harmful DLL that deploys a persistent loader and cryptocurrency miner. Users should download AI software only through official or IT-approved channels.
Google and its partners dismantled the XCPE botnet, which affected 1.3 million Android TV boxes. The network used compromised devices to relay web traffic and generate fraudulent advertising clicks. Google alleges that Videostrong planted firmware backdoors and has permanently prohibited the company from manufacturing certified devices. Affected boxes may still contain persistent firmware backdoors and should be disconnected pending clean firmware or replacement.
Cyberattacks threaten physical supply chains
Cybercriminals manipulated refrigeration systems at three French blood-service sites, although alarms allowed employees to intervene before blood products were destroyed. Similar incidents have spoiled food at European retailers, and the Pentagon is investigating freezer failures at six United States military bases. The cases demonstrate how attacks on operational technology can endanger critical medical and food supplies.
Industry shakeups
Robotics and AI infrastructure meet market reality
Agility Robotics made its Nasdaq debut under the ticker AGLT after merging with Churchill Capital XI. The transaction raised approximately $465 million before fees. The humanoid-robot maker is entering the public market despite reporting only $1.78 million in net sales for 2025 alongside a $140.2 million operating loss.
Voters across the United States approved restrictions on data centers, including bans, power caps, zoning limitations, and advisory measures. The results could delay planned AI infrastructure projects and may favor large cloud providers that already control suitable sites and possess the necessary permits.
Samsung says its high-bandwidth memory supply for next year is already sold out as AI demand strains manufacturing capacity. Prolonged shortages could increase hardware prices and have reportedly contributed to reduced production of budget Galaxy A smartphones.
AI faces copyright and surveillance scrutiny
An adviser to the European Union’s highest court said Google’s Gemini summaries could infringe publishers’ rights if they bypass paywalls or reproduce articles so closely that readers no longer need to visit the original source. The opinion is nonbinding and also concluded that there was no proof the complainant’s articles had been used to train Gemini.
The ACLU says police are using AI to analyze license-plate sightings and flag travel patterns considered suspicious. The practice has raised concerns about mass surveillance and traffic stops involving drivers who were not previously suspected of a specific crime.
Satellite operators pursue direct-to-phone networks
SpaceX is reportedly considering buying or leasing Boost Mobile to secure terrestrial network access for its telecommunications plans. A transaction could be complicated by Boost’s reliance on AT&T and T-Mobile infrastructure.
Meanwhile, the FCC approved Amazon Leo’s $11.5 billion acquisition of Globalstar. The deal supports Amazon’s plan for a direct-to-phone satellite constellation expected in 2028, although the company will still need additional permission before deploying the satellites.
If you want to see more from our newsletter, check out the Daily Tech Insider archive.