Liz Ticong

October updates from Palisade Research show advanced AI models still subvert clear shutdown commands as others comply, raising concerns about controllability.

Microsoft rolls out Copilot Mode in Edge, bringing chat, voice, and task automation into the browser with Actions and Journeys, plus clear privacy controls.

Microsoft’s AI chief, Mustafa Suleyman, says Copilot will never flirt or allow erotica, as safety-first updates add health guidance and a new Mico persona.

DeepSeek-OCR compresses long contexts up to 10× with 97% precision, scales to millions of pages per day, and is open source for more efficient LLMs.

ISACA’s 2026 poll shows AI threats on top: 59% fear AI attacks, 63% cite AI social engineering, and only 13% feel very prepared as boards elevate AI risk.

After the Dutch takeover of Nexperia, tensions erupted inside the chipmaker, drawing Chinese backlash and exposing Europe’s fragile tech supply chains.

Google and Livity’s Future Report finds AI routine for Europe’s teens, reshaping learning and creativity as they push for rules, labels and transparency.

OpenAI and the King Estate paused Sora generations of Dr. King after disrespectful deepfakes surfaced; estates can request removals as guardrails are tightened.

At Dreamforce in San Francisco, Salesforce showcased Henry, a humanoid robot on Agentforce, signaling a shift from cloud software to embodied AI in work.

Former Google CEO Eric Schmidt warns AI models can be hacked to remove guardrails, escalating proliferation risks as companies race to patch.

CDT survey finds teens bonding with chatbots: 19% report AI romance; 42% use AI for companionship or escape; 50% feel less connected to teachers.

AI-generated videos of deceased celebrities trigger backlash. Zelda Williams calls AI clips of her father “gross,” while a Tupac deepfake fuels wider unease.