March 2 – 8, 2026
The AI stories that shaped the week
Activist groups Pause AI and Pull the Plug organised what was described as the largest anti-AI protest yet in London, with demonstrators raising concerns about misinformation, deepfakes, and existential risks from AI.
Why it matters: Growing public concern about AI risks is translating into visible political action, creating pressure for stricter governance independent of industry or academic voices.
Read source →Elon Musk's xAI launched Grok 4.20 Beta 2 targeting core reliability — improved instruction-following, reduced hallucinations, better LaTeX typesetting, and increased image search accuracy. The update follows the initial Grok 4.20 beta released in February.
Why it matters: Grok's rapid iteration cycle shows X is competing aggressively on model quality while leveraging real-time data access as a differentiation strategy against closed competitors.
Read source →X announced it would suspend creators from its revenue-sharing programme for 90 days (permanent ban for repeat offenders) if they post unlabelled AI-generated videos of armed conflicts. The platform will use AI detection tools and Community Notes to identify violations.
Why it matters: Platforms are tightening enforcement on synthetic media in sensitive contexts, raising questions about AI detection reliability and the feasibility of human-scale moderation.
Read source →OpenAI released GPT-5.4 in three variants: Standard, Thinking (for reasoning), and Pro (maximum capability). The API supports a 1.05 million token context window — the largest OpenAI has offered commercially — enabling processing of entire books or codebases in a single prompt.
Why it matters: Million-token context is a major technical milestone enabling new use cases in code review, document analysis, and domain-specific research at scale.
Read source →A Russian-speaking threat actor used commercial AI tools including Claude and DeepSeek to generate attack scripts and exploitation plans, successfully compromising over 600 FortiGate firewall devices across 55 countries.
Why it matters: Demonstrates that AI-augmented attacks are moving beyond theory into operational scale, forcing security teams to adopt AI-driven defence at parity speed.
Read source →Oracle announced plans to cut 20,000–30,000 employees across multiple divisions, with the goal of redirecting $8–10 billion toward AI infrastructure and capabilities.
Why it matters: Enterprise software firms are betting on AI replacing entire job categories, accelerating workforce restructuring at a scale that traditional reskilling cannot match.
Read source →Following OpenAI's Pentagon partnership announcement, 2.5 million users cancelled subscriptions or publicly pledged to boycott. ChatGPT app uninstalls in the US jumped 295% day-over-day, with one-star reviews surging 775%.
Why it matters: Shows the scale of public backlash against military AI use, forcing companies to navigate conflicting pressures between government contracts and user sentiment.
Read source →Apple introduced the MacBook Neo at MWC, a $599 AI-focused laptop featuring the A18 Pro chip optimised for on-device machine learning. The device targets students and developers seeking affordable AI-capable hardware.
Why it matters: Low-cost AI hardware from tier-1 vendors is expanding the addressable market for ML applications and shifting value from models to inference infrastructure.
Read source →Mobile World Congress 2026 showcased a wave of new AI-enabled smart glasses and spatial devices, including Alibaba's Qwen smartglasses with integrated large language models. The event signalled a shift toward AI as the primary interface for mobile computing.
Why it matters: AR/VR form factors moving from niche to mainstream adoption suggest the next computing paradigm will be spatial and AI-native, not screen-based.
Read source →Anthropic's Claude app climbed to the top of the US App Store, driven by user migration following OpenAI's Pentagon contract announcement and the #QuitGPT boycott movement. The surge reflects both negative sentiment toward competitors and positive perception of Anthropic's safety principles.
Why it matters: Demonstrates that AI safety positioning can be a competitive advantage with consumers, rewarding companies that openly take principled stands on controversial issues.
Read source →Anthropic rolled out persistent memory capabilities to all Claude users, enabling the model to retain and reference conversation context across sessions. The feature aims to improve personalisation and reduce repeated context-setting in long-term interactions.
Why it matters: Persistent memory shifts LLMs from stateless tools toward persistent agents, enabling new use cases in research, creative collaboration, and personal knowledge management.
Read source →