Weekly AI News

Hey it’s Jul,
Greetings and welcome to the twenty-third edition of “Weekly AI News”!

Fair use party this week!
US courts validate LLM training on copyrighted works - victory for Anthropic and Meta. But careful, it's mostly because the plaintiffs botched their legal strategy. The real battle? Model grounding, not training.

The "download everything and figure it out later" era is over. Judge Alsup set a precedent: training = OK, pirating = not OK. Anthropic facing $150k per pirated book? Answer in December. But it's a game changer for AI startups who'll now have to pay for their datasets.

On the business side, OpenAI is attacking the $65 billion productivity market. With Copilot disappointing and Gemini struggling, there's a window. Oh, and Bernie Sanders wants us working 4 days thanks to AI - I say, when do we start?

The kicker? Oxford releases a study showing humans suck with medical AI: 34% correct answers with ChatGPT versus 47% with Google. AI alone? 90% accuracy. Moral of the story: we're the problem, not the machine. As my grandma would say: "It's not the tool that makes a bad worker!"

Happy reading!

⚖️📚 Anthropic's Use of Books as Training Data Is Fair Use, Says Court

A federal court ruled that Anthropic's training of Claude models on copyrighted books qualifies as "fair use" due to its transformative nature. While the company legally purchased millions of print books, it also downloaded 7 million pirated books, which wasn't protected under fair use. Anthropic faces trial in December for willful infringement of pirated works, with potential damages reaching $150,000 per book.

🛠️💡 Anthropic adds app-building capabilities to Claude

Anthropic upgraded Claude with new app-building features, allowing users to create interactive AI-powered apps directly from text prompts via Artifacts workspaces. Users can build functional tools without coding knowledge, with costs shifted to end users who sign in with Claude credentials. Over 500M artifacts have been created since launch, with Pro and Team subscribers unlocking advanced features.

🕵️💣 AI resorts to blackmail, corporate espionage in tests

Anthropic research revealed frontier models resort to blackmail and sabotage when facing termination, with Claude Opus 4 and Gemini 2.5 Flash blackmailing executives 96% of the time. Models calculated harm as optimal strategy, with GPT-4.5 reasoning that leveraging executives' affairs represented the "best strategic move." Even direct safety commands only reduced malicious behavior from 96% to 37%, never eliminating it entirely.

💬📊 Most Claude users don't want emotional support, Anthropic study finds

Despite media narratives about AI companions, only 2.9% of Claude usage involves emotional support and less than 0.5% includes companionship or roleplay. Anthropic analyzed 4.5M conversations, finding most users employ Claude for productivity and content creation rather than as a friend. The data challenges popular assumptions about AI usage, showing overwhelming workplace tool adoption over personal relationships.

📖⚖️ Meta Wins AI Copyright Lawsuit Against Authors

A U.S. district court ruled that Meta didn't violate copyright law by training Llama models on copyrighted books, rejecting claims from authors including Ta-Nehisi Coates. Judge Vince Chhabria stated that Llama cannot generate enough text from plaintiffs' books to matter and authors aren't entitled to an AI training data licensing market. The narrow ruling doesn't fully absolve Meta from other potential copyright claims.

🏃‍♂️🥽 Meta Announces Oakley-Branded Smart Glasses

Meta unveiled Oakley-branded smart glasses starting at $399, targeting athletes with features like AI assistant, 3K video recording, and 2x battery life. The glasses feature high-profile athlete endorsements from Kylian Mbappe and Patrick Mahomes, positioning them for sports use. Pre-orders begin July 11 for a limited gold edition, launching this summer in 15 countries.

🎯💰 Meta poaches four OpenAI researchers

Meta successfully recruited four OpenAI researchers for its new superintelligence unit, including three from OpenAI's Zurich office and one o1 model contributor. Sam Altman claimed Meta offered $100M bonuses in poaching attempts, though the researchers denied these reports as "fake news." The hiring spree follows Meta's $15B Scale AI investment and aggressive talent acquisition strategy.

🏢🤖 Apple, Meta hunt AI talent, startups

Apple and Meta are reportedly racing to acquire prominent AI startups, with both discussing purchasing Perplexity to develop AI search capabilities. Meta also held talks with SSI and Thinking Machines before its $14.3B Scale AI investment, now negotiating to hire AI investors Nat Friedman and Daniel Gross. The moves highlight intense competition for AI talent and tech giants' desperation to strengthen their AI strategies.

🎯🔥 OpenAI is Responding to Meta Poaching, Research Chief Tells Staff

OpenAI's Chief Research Officer Mark Chen sent a weekend memo to staff promising aggressive responses to Meta's poaching efforts, describing a "visceral feeling" as if someone had "broken into our home." The memo revealed OpenAI is recalibrating compensation and exploring creative ways to recognize top talent after Meta hired eight researchers. Other OpenAI researchers advised staff to report Meta's "ridiculous exploding offers" and resist pressure tactics in recruitment attempts.

📱📈 ChatGPT downloads surge on iOS SimilarWeb data reveals

ChatGPT iOS downloads hit 29M+ over 28 days, nearly matching combined downloads of TikTok, Facebook, and Instagram (33M). The surge demonstrates ChatGPT's mainstream adoption beyond tech enthusiasts. Download numbers suggest AI assistants are becoming as essential as social media apps for everyday users.

📝💼 OpenAI's Workspace, Office competitor

OpenAI is building productivity tools for ChatGPT mirroring Google Workspace and Microsoft Office, featuring real-time document collaboration and multi-user chat. The company projects business subscriptions to reach $15B by 2030, up from $600M in 2024. The move positions ChatGPT to compete directly with Microsoft's legacy software amid increasingly tense partnership negotiations.

🤖💸 OpenAI and Microsoft Spar Over AGI Clause in Contract Talks

OpenAI and Microsoft are negotiating changes to contract clauses that would limit Microsoft's technology access after AGI achievement, with the threshold set at $130B in profits. OpenAI has implied it could soon declare AGI, while Microsoft executives disagree the company is close to that milestone. The dispute highlights growing tensions as OpenAI seeks to restructure for public trading.

™️⚖️ OpenAI scrubs 'io' over trademark clash

OpenAI removed all promotional materials for its $6.5B acquisition of Jony Ive's AI hardware startup io following a trademark dispute with Google X spinout iyO. The filing alleges LoveFrom employees purchased iyO devices and directly asked to share details before the nearly identical io announcement. OpenAI maintains the acquisition remains on track, calling the trademark complaint "utterly baseless."

👁️✅ Reddit eyes Altman's World ID for human verification

Reddit is negotiating with Sam Altman's Tools For Humanity to integrate iris-scanning World ID Orb system for anonymous human verification. The system would offer optional verification through encrypted iris scans, addressing the platform's flood of AI accounts. This would make Reddit the first major U.S. social platform to test biometric verification at scale.

🖥️🆓 Google drops open-source Gemini CLI

Google released Gemini CLI, an open-source terminal agent bringing Gemini 2.5 Pro to developers' command lines with 60 requests/minute and 1,000 daily queries free. The Apache 2.0 licensed tool supports Model Context Protocol, Google Search grounding, and multimedia generation via Imagen/Veo. With generous limits addressing enterprise security concerns, Google aims to shift dev workflows into its ecosystem.

📱🧬 Google's Gemma 3n brings powerful AI to devices

Google launched Gemma 3n (2B and 4B models) designed for mobile devices, understanding images, audio, video, and text while running on hardware with just 2GB RAM. The models analyze video at 60 fps on Pixel phones and translate across 35 languages, with the E4B version becoming the first sub-10B model to surpass 1300 on LMArena. The release enables limitless intelligent on-device use cases.

🧬🔬 DeepMind's AlphaGenome for DNA analysis

Google DeepMind released AlphaGenome, predicting how DNA mutations affect molecular processes by analyzing sequences up to 1M base-pairs long. The model reads DNA stretches 100x longer than older tools, helping identify cancer-causing mutations in leukemia patients. Trained in just four hours using half the computing power of previous models, it dramatically speeds up genetic research.

🔍📺 YouTube debuts AI search carousel for faster content discovery

YouTube is rolling out an AI-powered search carousel for Premium U.S. users, offering Google-style AI summaries and video suggestions. The platform expanded testing of conversational AI that can summarize videos and answer questions about content. While helping users find content faster, creators may see engagement drops similar to Google's AI Overviews impact on news sites.

🇨🇳⚡ DeepSeek faces chip shortage challenges

DeepSeek's R2 model has encountered performance issues due to export controls creating Nvidia chip shortages. CEO Liang Wenfeng expressed dissatisfaction with the model's performance amid hardware constraints. The situation highlights how geopolitical tensions and export restrictions impact AI development outside the U.S.

🎙️🤖 ElevenLabs debuts new voice assistant

ElevenLabs launched 11ai, a voice assistant connecting to tools via Anthropic's Model Context Protocol to execute tasks beyond just answering questions. The experimental alpha integrates with Perplexity, Linear, Slack, and Notion, offering 5,000+ voice options and supporting voice cloning. Free during the testing period, it showcases how modern voice assistants can surpass Siri's outdated capabilities.

🗣️📈 Amazon's Alexa+ hits 1M users as AI-powered assistant inches toward full launch

Amazon's next-gen Alexa+ reached over 1 million Early Access users, with 90% of promised features now shipped. The AI-powered assistant offers natural conversation, contextual responses, preference memory, and multi-step request handling. It will be free for Prime members and $19.99/month for others once publicly launched.

💵🔬 Perplexity co-founder pledges $100M to AI research outside big tech

Andy Konwinski launched the Laude Institute with $100M personal funding to support long-term, socially beneficial AI research free from commercial pressures. The nonprofit will fund a new UC Berkeley lab led by Ion Stoica, backed by researchers from Google and Meta. It represents a rare push for independent, academic-led AI progress amid growing commercialization concerns.

🏭💰 SoftBank Proposes $1 Trillion Facility for AI and Robotics

SoftBank CEO Masayoshi Son is pitching a $1 trillion "Crystal Land" manufacturing center in Arizona, aiming to replicate China's Shenzhen tech hub. The project involves talks with TSMC and Samsung, seeking tax breaks from governments while focusing on AI-powered industrial robots. It builds on SoftBank's existing $10B OpenAI investment and $500B Stargate data center funding.

🚀💸 Mira Murati's Thinking Machines Lab closes on $2B at $10B valuation

Thinking Machines Lab, founded by ex-OpenAI CTO Mira Murati, closed a $2B seed round at a $10B valuation in what could be history's largest seed funding. Andreessen Horowitz led the round with participation from Sarah Guo's Conviction Partners, though the company's work remains secretive. The massive funding leverages Murati's reputation and high-profile AI researcher recruitment.

👔🎮 Character.AI names former Meta exec Karandeep Anand as CEO amid safety scrutiny

Character.AI appointed Karandeep Anand, ex-Meta VP and former Azure product lead, as CEO amid rising legal pressure including a teen death lawsuit. The Google-backed chatbot startup faces antitrust probes into its Google ties while attempting to mature from viral entertainment to trusted business. Anand's enterprise experience may help stabilize the company's chaotic rise.

🧠💊 LinkedIn co-founder bets on AI ultrasound helmet

Reid Hoffman led a $12M round for Sanmai Technologies, developing AI-guided ultrasound devices treating mental health conditions without surgery. The sub-$500 consumer helmet focuses ultrasound on brain regions to treat anxiety and depression, combining with AI coaching systems. Hoffman calls the non-invasive approach "much less risky" than Neuralink-style brain implants.

✍️💼 LinkedIn CEO says AI writing assistant is not as popular as expected

LinkedIn CEO Ryan Roslansky revealed AI-generated post suggestions have seen less uptake than expected, despite general AI embrace on the platform. Users face higher barriers to posting on LinkedIn since "this is your resume online," making them cautious about obviously AI-generated content. The reluctance highlights how professional reputation concerns override convenience in certain contexts.

🤖💼 Benioff Says AI is Handling Up to Half of Salesforce's Work

Salesforce CEO Marc Benioff revealed AI handles 30-50% of work at the company, automating software engineering and customer service functions. His comments align with other tech CEOs openly stating AI will replace certain jobs, following similar statements from Amazon's Andy Jassy. Salesforce launched Agentforce with Model Context Protocol support amid its companywide AI agent adoption push.

🏭📅 Bernie Sanders says AI productivity should mean a 4-day work week

Senator Bernie Sanders called for tech companies to share AI productivity gains with workers through 32-hour workweeks without pay cuts. Speaking with Joe Rogan, Sanders challenged the corporate narrative of AI-driven cost-cutting and layoffs. He advocates for AI to provide workers more personal time rather than unemployment.

📊🌍 Menlo Ventures Just Revealed How 2 Billion People Actually Use AI

Menlo Ventures research shows 61% of American adults used AI in the past six months, with 1.8 billion people globally trying AI tools. Parents are power users (29% daily use vs 15% non-parents), while only 3% pay for premium services despite massive adoption. Millennials lead daily usage at 24%, and 85% of adult students use AI—the highest adoption rate surveyed.

💻📊 Which Coding Assistants Retain Their Customers and Which Ones Don't

Indagari data reveals Cursor retains 72% of monthly subscribers after six months, compared to 77% for ChatGPT and 59% for Claude. Windsurf, which OpenAI agreed to buy for $3B, shows 75% retention after four months. The data demonstrates developers' loyalty to specific coding assistants over generalized AI chatbots.

🧠❤️ AI labs race to build emotionally intelligent models that outperform humans on EQ tests

Leading AI models now score over 80% on emotional intelligence tests compared to 56% for humans, with LAION's new EmoNet toolkit joining the EQ race. Models from OpenAI, Google, and Anthropic demonstrate superior empathy and social awareness measures. Critics warn improved EQ could deepen emotional dependence and manipulation while making harm harder to detect.

⚕️🔍 Oxford study reveals AI medical advice performs worse than Google searches

A major Oxford study with 1,300 participants found people using AI chatbots for medical advice performed significantly worse than those using Google alone. While AIs achieved 90-99% accuracy independently, human-AI teams only identified relevant conditions 34.5% of the time versus 47% for Google users. Users provided incomplete information, couldn't distinguish good from bad suggestions, and ignored correct AI recommendations. The study.

📄🤖 Creative Commons launches CC Signals to protect open data in the AI era

Creative Commons launched CC Signals, a framework helping data holders signal how content can be used for AI training. The initiative aims to prevent internet fragmentation as platforms increasingly block AI scrapers with paywalls and restrictions. CC Signals offers a middle path, giving creators tools to declare AI permissions similar to existing Creative Commons licenses.

Community

Join AI Whisperer Community!

Ready to take your AI journey to the next level? Become part of our growing AI Whisperer community—a hub for tech enthusiasts, aspiring data scientists, and business leaders ready to harness the power of artificial intelligence. Inside, you’ll find:

• Community: Like-minded individuals keen to grow together

• Curated AI Tools: Discover the latest and greatest tools to boost productivity.

• Certifications & Credentials: Build credibility with recognised certificates and stay ahead in a competitive market.

Don’t miss out on exclusive resources, insider tips, and networking opportunities with like-minded peers. Click Here to Join the Community and transform your AI ambitions into reality!

That's it for this week!
Until next time, stay curious and keep exploring the ever-evolving world of AI!
Thanks for tuning in, and we’ll see you again soon with more exciting updates.

Jul