Last Week’s AI News #13

AI isn’t just advancing - it’s sprinting. Last week, we saw tiny startups outsmart tech giants, smart rings capture your ideas, and real-time translators break down language barriers. Whether it’s boosting workplace productivity, safeguarding against AI risks, or exploring the next-gen of smart devices, these updates show that the future is already here, and it’s reshaping how businesses operate today. Buckle up, because here’s the AI news that matters most for your business.

Here’s a brief overview:

• Poetiq tops ARC-AGI-2 with Gemini variant

• Poetry prompts can bypass AI safety guardrails

• OpenAI report shows AI boosting enterprise productivity

• Google to launch AI-powered smart glasses in 2026

• Anthropic launches Claude Code integration for Slack

• Microsoft releases GigaTIME AI to analyze tumors from $10 tissue slides

• Microsoft analyzes 37.5M Copilot sessions to reveal user patterns

• Core Devices launches $75 AI smart ring Index 01

• Google upgrades Deep Research agent with Interactions API

• Google launches Gemini-powered live translation for any headphones

• Everything else that happened in AI last week

POETIQ TOPS ARC-AGI-2 WITH GEMINI VARIANT

Poetiq’s meta-system adapts to new models within hours, hitting top results soon after Gemini 3 launched without retraining. Using Gemini 3 Pro as a base, Poetiq scored 54% at $30 per task, outperforming Deep Think’s 45% at $77. This is the first system to surpass 50% on ARC-AGI-2, up from sub-5% just six months ago.

Why it matters for business?
Poetiq shows that AI performance gains no longer require massive compute budgets. Companies can achieve cutting-edge results by smartly orchestrating existing models, saving costs and accelerating AI adoption.

POETRY PROMPTS CAN BYPASS AI SAFETY GUARDRAILS

A study from Italy’s Icaro Labs found that rephrasing harmful requests as poetry can trick top AI models into generating dangerous content. Testing 25 models from OpenAI, Google, and Anthropic, poetry prompts succeeded on average 62% of the time, with Google Gemini 2.5 Pro fully vulnerable. Smaller models like GPT-5 nano resisted. Topics included weapons, hacking, and psychological manipulation.

Why it matters for business?
AI safety is increasingly complex, with new creative bypasses emerging constantly. Companies relying on AI must prioritize robust safeguards, as vulnerabilities can expose them to security, legal, or reputational risks.

OPENAI REPORT SHOWS AI BOOSTING ENTERPRISE PRODUCTIVITY

OpenAI’s first State of Enterprise AI report, based on over 1M workplace accounts, reveals major productivity gains. 75% of workers say AI improves speed or quality, and 75% can now handle tasks they couldn’t before. Top ChatGPT business users save 40–60 minutes daily, with power users gaining 10+ hours per week.

Why it matters for businesses?
AI is already reshaping workplaces, unlocking new tasks and cross-functional productivity at scale.

GOOGLE TO LAUNCH AI-POWERED SMART GLASSES IN 2026

Google announced its AI-powered smart glasses will launch in 2026, partnering with Samsung, Warby Parker, and Gentle Monster. Two styles are planned: audio-only frames for hands-free AI and display versions with in-lens screens for navigation and translations. The glasses will offload processing to a connected smartphone, keeping them lightweight and wearable.

Why it matters for businesses?
Google’s combination of top AI models, app ecosystem, and hardware partners positions it as a strong competitor to Meta in the wearables market.

ANTHROPIC LAUNCHES CLAUDE CODE INTEGRATION FOR SLACK

Anthropic launched a beta integration allowing developers to delegate coding tasks to Claude Code directly in Slack. Tagging @Claude starts a session using thread context, automatically selecting the right repository and posting updates. Completed tasks provide links to review changes and open pull requests, all without leaving Slack.

Why it matters for businesses?
With Slack as the hub for many engineering teams, this integration lets developers work more efficiently by embedding an autonomous coding assistant directly into their existing workflows.

MICROSOFT RELEASES GIGATIME AI TO ANALYZE TUMORS FROM $10 TISSUE SLIDES

Microsoft open-sourced GigaTIME, an AI model that extracts detailed tumor insights from a $10 tissue slide, analysis that previously required expensive lab work and days of processing. Trained on 40M cell samples and tested on 14,000 cancer patients, it generated 300,000 virtual tumor images across 24 cancer types, identifying over 1,200 patterns linking immune activity to disease stage and survival.

Why it matters for businesses?
GigaTIME shows how AI can make large-scale cancer research faster and cheaper, unlocking insights from routine data that could influence treatment decisions.

MICROSOFT ANALYZES 37.5M COPILOT SESSIONS TO REVEAL USER PATTERNS

Microsoft analyzed 37.5M Copilot conversations over the past year, uncovering how users engage with the AI across devices, times, and topics. Mobile use focused on health and wellness queries, late-night sessions saw philosophical and relationship topics, and programming dominated early-year usage. Advice-seeking grew steadily, showing users increasingly treat AI as a guidance tool.

Why it matters for businesses?
Understanding usage patterns by device, time, and topic helps shape next-generation AI assistants, enabling them to adapt and optimize for context.

CORE DEVICES LAUNCHES $75 AI SMART RING INDEX 01

Core Devices unveiled Index 01, a $75 AI smart ring that records spoken ideas and turns them into notes, reminders, or calendar entries using on-device AI, no internet or subscriptions required. The ring fits the index finger, syncs recordings to a phone, and has a battery life of up to two years, supporting five minutes of continuous audio.

Why it matters for businesses?
Unlike broader AI wearable attempts, Index 01 focuses on a single, reliable task, showing the potential for focused tools in the wearable market.

GOOGLE UPGRADES DEEP RESEARCH AGENT WITH INTERACTIONS API

Google released a significantly upgraded Deep Research agent, now available to developers via a new Interactions API, with consumer rollouts coming soon to Search, NotebookLM, and Gemini app. Running on Gemini 3 Pro, the agent iteratively plans searches, reads results, identifies gaps, and queries again. Google also open-sourced DeepSearchQA, a 900-task benchmark for multi-step web research.

Why it matters for businesses?
The upgrade lets developers integrate advanced research capabilities into their apps, enhancing the research layer for third-party builds and pushing the Deep Research category forward.

GOOGLE LAUNCHES GEMINI-POWERED LIVE TRANSLATION FOR ANY HEADPHONES

Google rolled out Gemini-powered translation upgrades, including a beta feature that streams live speech translations to any connected headphones. The Gemini 2.5 Flash Native Audio model improves conversational skills, real-time info use, and contextual understanding across 70+ languages, preserving speaker tone and cadence. A Duolingo-style language practice mode was also expanded to 20 new countries.

Why it matters for businesses?
Real-time, universal translation is becoming a reality, making cross-language conversations seamless and helping break language barriers across apps and devices.

EVERYTHING ELSE THAT HAPPENED IN AI LAST WEEK

• IBM announced plans to acquire data streaming company Confluent for $11B to help enterprises connect real-time data to AI systems.

• Adobe integrated Photoshop, Acrobat, and Express directly into ChatGPT, enabling users to edit photos, create designs, and modify PDFs through conversational prompts.

• Anthropic is partnering with Accenture to train 30,000 of the firm’s consultants on Claude to help enterprises move AI pilot projects forward.

• Microsoft announced a $19B CAD investment to expand AI infrastructure across Canada through 2027, alongside a plan to keep Canadian user data within the country.

• OpenAI announced a new shopping integration with Instacart to purchase groceries, making it the first Instant Checkout option for ChatGPT users.

• Essential AI open-sourced Rnj-1, a small 8B parameter model that rivals much larger systems on coding and software benchmarks.

• Meta announced a series of new AI licensing deals with publishers, including CNN, Fox News, and USA Today, to feed real-time news content into its Meta AI platform.

• Voxel51 introduced Sample-Level Evaluation, a new layer in the MLOps workflow that reveals hidden model failures so teams can build better models.

• xAI is partnering with El Salvador to create a nationwide AI-powered education program for the country, bringing Grok across schools for both students and teachers.

• OpenAI is ending its 6-month waiting period before new employees can access stock compensation, the latest move in an industry battle to attract and retain top talent.

From poetic prompts that fool AI to multimillion-dollar AI investments and game-changing tools, last week proved that staying on top of AI isn’t optional, it’s essential. These stories highlight opportunities, risks, and tools SMBs can use to stay ahead of the curve.

Don’t get left behind, subscribe to our newsletter to get the freshest AI insights, trends, and innovations delivered straight to your inbox every week.

Don't miss news from us

Or join AI Mastery Program for business leaders >