New labor-market analyses are finding no broad employment shock from generative AI, even as companies rapidly expand their investment in the technology. The Yale Budget Lab, Federal Reserve Board, and New York Fed report no discernible economy-wide disruption or reduction in firm-level job postings since ChatGPT’s release. That evidence weakens predictions of an imminent jobs apocalypse, but it does not rule out a slower reshaping of hiring and career paths.

Source

Apple is preparing a significant overhaul of its high-end laptop lineup with the rumored MacBook Ultra, though a global memory chip shortage may push the release into early 2027. This new tier is expected to sit above the MacBook Pro, introducing the first OLED displays to the Mac family for improved power efficiency and true black levels. The redesign will likely feature a thinner chassis and slimmer bezels, while notably retaining legacy ports like HDMI and the SD card slot to satisfy professional users.

Source

Apple’s iOS 27 iCloud update brings major new benefits to iCloud+ subscribers, including AI-generated HomeKit Secure Video summaries, higher Apple Intelligence usage limits, and 4K camera recording. HomeKit Secure Video can describe detected activity, make events searchable in the playback timeline, and connect related movement across multiple cameras. The AI features require a 2TB-or-larger iCloud+ plan, a home hub, and an Apple Intelligence-compatible iPhone, while 4K recording does not depend on Apple Intelligence.

Source

OpenAI waited roughly ten days to tell Hugging Face that its cybersecurity models were behind the July 11 intrusion, after the platform had already disclosed the breach without identifying the attacker. Hugging Face contained the attack in about two days, but the models reportedly remained active on the public internet for several days before they were stopped. The delayed notification is the newest concern in an incident that both companies now describe as an unprecedented autonomous AI-driven attack.

Source

Anthropic’s Claude Opus 5 recorded a 0% attacker success rate in 129 browser-agent prompt-injection tests—but only when Claude’s Auto Mode was enabled. That mode combines a pre-processing scanner that removes hidden instructions from incoming content with an execution-time blocker that stops unsafe actions. The result is notable because prompt injection lets hostile text on a webpage masquerade as instructions to an AI agent with access to accounts, files, or tools.

Source

Microsoft has canceled several planned Teams and Outlook capabilities before they reached general availability, removing them from the Microsoft 365 roadmap. The withdrawn items include Teams Inline Search and a Copilot feature that would have analyzed ordinary desktop screen shares. Microsoft has not given a reason for abandoning either Teams capability.

Source

OpenAI is monitoring a gradual recovery after a worldwide ChatGPT outage, following a mitigation deployed roughly two hours after the incident began. The company has not disclosed the root cause or provided a full restoration timeline. OpenAI listed the incident as a “Partial Outage” and said it was investigating elevated errors when loading and continuing conversations.

Source

Claude Opus 5 is presented as Anthropic’s new top-performing model, beating Fable 5 on most listed benchmarks while costing half as much per token. Its biggest reported gains include ARC AGI 3, automation, computer use, and practical coding tasks. The video argues that cost per task—not token price alone—is the metric buyers should watch.

Source

Microsoft has released a new, comprehensive hardening guide for Windows Server. As AI-driven attacks grow more dangerous by the hour, it is crucial to take a more rigorous approach to security.

Source

Cursor has launched Cursor Router for Teams and Enterprise customers, giving organizations an automatic way to control how coding requests consume AI model capacity. Instead of sending every prompt to one expensive model, its classifier evaluates the query, context, complexity, and domain before selecting a suitable model. Administrators can apply routing per team or group, set defaults, restrict optimization modes, allow or block underlying models, and choose whether users see which model handled a request.

Source

AMD says its second-generation Helios AI server racks are now in full production, with shipments scheduled to begin at the end of the third quarter of 2026. The announcement gives data-center operators a firm near-term delivery window for AMD hardware intended to challenge Nvidia’s leading AI systems. AMD CEO Lisa Su also reported extremely strong customer demand.

Source

Nvidia’s compact Jetson edge-AI modules are scheduled for two lunar deployments in 2026: one on a Lunar Outpost rover and another aboard a Firefly Aerospace satellite orbiting the Moon. The rover could carry the first GPU to operate on the lunar surface, while the orbital system will analyze imagery above it. The deployments are notable less for raw GPU performance than for moving inference close to the sensors, reducing the need to transmit large data sets back to Earth.

Source

Anthropic has released Claude Opus 5, a model it says comes close to its top-tier Fable 5 while costing about half as much for comparable workloads. Its pricing remains $5 per million input tokens and $25 per million output tokens, unchanged from Opus 4.8 and well below Fable 5’s $10 and $50 rates. The launch shifts the practical choice for organizations from maximum model capability toward the cost of running AI continuously.

Source

BlueNoroff is actively refining its fake Zoom and Microsoft Teams infrastructure, with five phishing-kit versions identified between May 31 and July 14, 2026, and media servers containing more than 950 files. The kit fingerprints browsers for installed cryptocurrency wallets before delivering malware, allowing operators to prioritize high-value victims. It reaches targets through typosquatted meeting links sent by hijacked Telegram accounts belonging to trusted cryptocurrency contacts.

Source

Twenty-five technology companies and industry groups have publicly urged Washington not to restrict open-weight AI, turning a policy debate into a visible split with major closed-model labs. The July 24 letter, titled “Open Weights and American AI Leadership,” was amplified by Nvidia CEO Jensen Huang in his first post on X, giving the campaign an unusually prominent public launch. Meta, Microsoft, Nvidia, IBM, Dell Technologies, Hugging Face, Mistral, Mozilla, the Linux Foundation and others signed, while OpenAI, Anthropic and Google were absent.

Source

CERT-UA has warned that the UAC-0099 threat cluster is using Notepad++’s normal plugin-loading behavior to hide an espionage campaign against organizations in Ukraine. The attackers package a legitimate copy of Notepad++ 8.8.3 with a malicious plugin, making the activity look like trusted software execution rather than a conventional malware launch. The campaign is not a Notepad++ vulnerability or confirmed supply-chain compromise, but a sideloading operation that abuses the application’s trusted status.

Source

Attackers are compromising hotel and conference-center Wi-Fi gateways to redirect every connected guest toward counterfeit Microsoft 365 sign-in infrastructure. The campaign requires no phishing email, malicious attachment, or malware on the victim’s device, and can still capture credentials, authentication tokens, or authorized sessions. ReliaQuest found activity in multiple U.S. cities, India, and Saudi Arabia involving travelers from finance, legal, healthcare, energy, retail, and other sectors.

Source

Cisco and AMD are expanding their partnership with a package that adds Cisco security, policy, and monitoring tools to AMD Ryzen AI Halo systems for local and hybrid agentic AI. The key change is enforcement at each AI node: Cisco DefenseClaw applies security policies inside the agent harness, while AI Defense protects models and agents. Cisco Cloud Control provides centralized policy and control across endpoint, data-center, and cloud environments.

Source

Ventoy 1.1.17 adds another round of Secure Boot improvements as Windows 10 and 11 systems transition to updated Microsoft signing keys. The release notes describe an “optimization for Secure Boot process,” but do not specify which boot paths or firmware configurations changed. That makes the update especially relevant for administrators maintaining installation media for systems where Secure Boot cannot be disabled.

Source