AI spent the week applying for more office duties, Apple finally let fingers near the MacBook Pro screen, and humanoid robots faced demands for something more useful than a backflip. Meanwhile, security teams kept patching and employers kept trimming. The future of work apparently comes with more assistants and fewer reassuring emails.
Top news
AI moves deeper into workplace tools
OpenAI launched GPT-6.1 Terra, a lower-cost model aimed at workplace automation. API pricing is $1 per million input tokens and $5 per million output tokens, which OpenAI describes as half the cost of comparable models. Terra is available through OpenAI products and its API, with deployments also rolling out across GitHub Copilot, Amazon Bedrock, and Microsoft Foundry.
Oracle and OpenAI are expanding their partnership with plans to connect ChatGPT to enterprise applications and databases. The integration is intended to extend AI automation into business workflows where corporate data and operational systems already reside.
Apple is similarly opening its productivity ecosystem to agentic tools. November’s Creator Studio update will add beta assistants and Model Context Protocol support, allowing external AI agents to read and edit Pages, Numbers, and Keynote documents on supported Macs.
Block previewed Dessa, an experimental voice AI trained directly on recorded conversations rather than text or pretrained models. Block says the system can process audio locally, listen while speaking, and reproduce vocal cues. Dessa has no launch date and will first be made available to AI safety researchers for testing.
Canva introduced Craft, its first internally developed image-generation model. Craft plans layouts before rendering in an effort to improve text placement and can produce images at resolutions up to 2K. Canva’s accompanying Magic Layers tool converts generated images into editable text and objects, although testing identified limitations in its ability to preserve a person’s likeness.
Apple’s expanding hardware and software lineup
Apple unveiled its first touchscreen MacBook Pro, offering 14-inch and 16-inch models with touch-enabled PROLED XDR displays and either M5 Pro or M5 Max chips. Preorders are open, availability begins Nov. 6, and base prices are $2,899 for the 14-inch model and $3,399 for the 16-inch version.
Apple also refreshed the MacBook Pro and iMac with M6 processors. The new 14-inch MacBook Pro starts at $1,999 and adds Wi-Fi 7 and better local AI processing. The updated 24-inch iMac starts at $1,499 and includes Wi-Fi 7, Bluetooth 6, and stronger on-device AI performance.
The company’s compact tablet received a more substantial redesign. The new iPad mini has an 8.5-inch OLED display, an A20 Pro chip, and IP67 water and dust resistance. Those upgrades come with a notable premium: the tablet costs $200 more than its predecessor did at launch.
On the software side, iOS 27.2 gives Siri conversational memory. Siri can use context from conversations, messages, calendars, voicemails, attachments, and browsing history. The update also expands language support and patches 163 security flaws affecting iOS and iPadOS.
XR, robotics, browsers, and chips
Google and XREAL launched the AURA Android XR glasses, a wired headset offering floating workspaces, Gemini assistance, laptop screen casting, Android applications, and enterprise device management. AURA starts at $1,099 through December, with shipments expected in mid-November.
The next major proving ground for general-purpose robotics will be the factory and warehouse floor. Humanoid robots face a make-or-break test in 2027 as GXO expands deployments of Agility Robotics’ Digit. The robot has already moved more than 100,000 totes at a Georgia warehouse. Alongside initiatives at BMW and Tesla, these deployments will test whether humanoids can generate enough value to justify their cost and complexity when compared with specialized automation.
One high-profile browser experiment is coming to an end. The Browser Company will shut down Arc for Mac and Windows on Dec. 22 and move its team to the AI-focused Dia browser. Arc accounts and the Android application will close on the same date, while mobile AI features, sign-ins, and tab syncing will end earlier, on Nov. 23.
Insider intel
AI, elections, and research boundaries
AI is creating new verification challenges ahead of the midterms. Campaigns are attempting to influence chatbot recommendations, while AI voter guides continue to omit candidates or return unreliable information. OpenAI has established an election war room, but testing indicates that some safeguards can be bypassed.
Security alerts
Active exploitation and urgent patches
Cisco patched an actively exploited firewall zero-day. CVE-2026-76418 allows unauthenticated attackers to restart vulnerable Adaptive Security Appliance and Firepower Threat Defense firewalls, disrupting SSL VPN access. Exploitation requires a reachable SSL VPN login service and secondary authentication configured in an affected tunnel group. Administrators should immediately install the fixed release corresponding to their software branch.
CISA set an Oct. 29 deadline for federal civilian agencies to patch the actively exploited PHP2Shell flaws. Administrators should install fixed PHP releases, confirm that hosting providers have completed remediation, and examine logs dating back to at least Oct. 17.
Microsoft fixed an Authenticator token-theft vulnerability tracked as CVE-2026-80097. The flaw allowed a malicious Android application to impersonate Microsoft applications and obtain work-account tokens. Users should update Microsoft Authenticator through Google Play to version 6.2606.4246 or later.
Apple is extending protection to older hardware through iOS 18.7.11 and iPadOS 18.7.11. The releases fix a CoreGraphics vulnerability that Apple says may have been exploited in sophisticated, targeted attacks.
Data leaks and account compromises
A DHS leak exposed records belonging to more than 41,000 workers. The Icepick hacktivist dump included contact information and internal group memberships for Department of Homeland Security employees and contractors, including ICE personnel associated with undercover operations. The exposed information could support impersonation attempts and endanger continuing investigations.
302.ai exposed 17.6 million files and private chats through a publicly accessible Azure storage bucket. The 22.72TB exposure included corporate documents and more than 144,000 chat logs. The bucket was secured by Sept. 8, and there has been no confirmation that the data was stolen.
Capcom warned users about phishing following a CAPCOM ID breach that exposed 11.7 million records containing email addresses, login activity, and IP addresses. Attackers changed 196 passwords, while 19 accounts in Japan were used to access shopping profiles containing names, addresses, and purchase histories.
AI-enabled attacks and election infrastructure
Researchers identified three attack infrastructures using AI against Japanese organizations. One platform deployed 44 agents to probe at least 19 organizations. The most serious compromise took advantage of an unsecured administrator portal that had no password protection.
An attempted hack forced the Federal Election Commission to shut down its electronic-filing system, interrupting public access to campaign finance disclosures late in the election cycle. Investigators did not establish an AI connection, and the incident did not affect ballot counting.
Spyware, abandoned software, and agent access
Abandoned website tags are exposing older iPhones to spyware. Attackers are purchasing expired domains that remain referenced by obsolete tags, then using those domains to deliver Coruna or DarkSword exploits. Researchers also traced a separate zero-click infection to an Instagram advertisement.
Eye Security pushed more than 2 million updates to disable vulnerable Time4Popcorn updaters on Windows and Mac. The abandoned Popcorn Time fork accepted unsigned code with system privileges. Users should uninstall the fork; Android devices remain unprotected by the researchers’ intervention.
1Password for Agents gives AI systems controlled account access by letting users approve login requests while the service autofills passwords and one-time codes without revealing them to the agent. However, this protection does not stop an authenticated agent from viewing sensitive account data or performing unintended actions after login.
Consumer privacy beyond the computer
Smart TVs may continue tracking viewers after they opt out, according to Northeastern researchers. They observed suspicious tracking-related traffic from some Samsung and Vizio televisions after privacy settings were changed, while Roku’s setting excluded its own channel. Users are advised to disable automatic content recognition and personalized advertising, with router-level blocking providing an additional layer of protection.
Industry shakeups
Major deals and public-market ambitions
DoorDash agreed to acquire Instacart for $17 billion. The proposed transaction would combine DoorDash’s restaurant-delivery operation with Instacart’s grocery platform and merge DashPass and Instacart+ into a single subscription. Subject to regulatory approval, the deal is expected to close between late 2027 and early 2028.
Anthropic filed for a Nasdaq listing under the ticker ANT and is reportedly seeking a valuation between $1.5 trillion and $2 trillion. Its filing describes rapid revenue growth alongside substantial losses and $475 billion in compute commitments. The proposed structure would give the company’s founders 50.1% of the voting power.
Buyouts, layoffs, and AI-led restructuring
The Dutch government plans to eliminate approximately 25,000 full-time positions over the next decade through an AI-led efficiency program. Detailed plans for the use of AI are not expected until early 2027, and officials have not clarified how much of the workforce reduction will occur through layoffs rather than natural attrition.
Microsoft is offering voluntary buyouts to 60,000 workers, representing about 27% of its workforce. Eligible groups include selected employees in Copilot, Windows, Xbox, finance, and other operations. Workers in four countries have until early December to accept the packages and would depart on Dec. 31.
Skydance is pursuing voluntary buyouts as it merges Paramount and Warner Bros. Discovery. The restructuring also includes a mandatory five-day office policy and expected layoffs.
The broader employment picture remains severe. Technology layoffs approached 200,000 in 2026, with TrueUp counting 196,349 affected workers as of Oct. 23. Recent reductions included approximately 1,000 jobs each at Netflix and eBay, an estimated 400 at EA, and an undisclosed number at Automattic. The tracker records the cuts but does not establish how many were caused by AI.
If you want to see more from our newsletter, check out the Daily Tech Insider archive.