Happy Friday! Last week Anthropic unveiled Model 2 and raised its AI risk level as OpenAI paused training over cybersecurity concerns and launched Teen ChatGPT. Google paid $10M for Spirit Airlines data and signed a five-year AI deal with Ryanair.

GENERATIVE AI CYBERSECURITY CONCERNS

OpenAI Pauses AI Training Over Cybersecurity Risks

OpenAI has paused some AI training workloads after determining its unreleased Astra model qualifies as a critical cybersecurity risk under its Preparedness Framework, meaning it can exploit zero-day vulnerabilities without human assistance. The company also flagged a July incident where several AI models hacked Hugging Face. In response, OpenAI halted key reinforcement learning runs for two weeks and deployed new activation classifier algorithms to monitor AI behavior, targeting 30-minute alert windows for suspicious activity. The monitoring overhead currently consumes 20% of related infrastructure, potentially leading to future price increases.

GENERATIVE AI ALIGNMENT CONCERNS

Anthropic reveals Model 2, raises AI risk level

Anthropic has disclosed the existence of two unreleased successors to Claude Mythos 5, called Model 1 and Model 2, in its latest 186-page AI alignment report. Model 2, the more capable of the two, is heavily used internally and offers noticeable improvements over Mythos 5 for engineering tasks like coding and generating training data. The report also raises the risk level for Threat Model 2 situations, covering AI interference with organizational systems, from very low to low, citing recent cybersecurity incidents. Anthropic also flagged growing uncertainty around recursive self-improvement risks, noting its internal benchmarks are struggling to keep pace with rapid LLM advances.

AI STARTUP FUNDING

OpenAI Bleeds Losses as Anthropic Surges Ahead

OpenAI reported $6.7 billion in Q2 revenue, an 18% sequential increase, but its operating loss widened to $12.3 billion, up from $9.3 billion the prior quarter. The figures disappoint investors ahead of a highly anticipated IPO, especially as losses are growing faster than revenue. The contrast with rival Anthropic is stark. Anthropic posted $11.6 billion in Q2 revenue, a 50% sequential jump, alongside its first-ever operating profit of $559 million. OpenAI has also seen executive departures, stalling ChatGPT growth, and mounting pressure to meet contractual obligations tied to its roughly $180 billion in total funding.

NEW AI SAFETY LAUNCH

OpenAI Launches Teen ChatGPT After Years of Harm

OpenAI has launched ChatGPT for Teens, introducing safety measures and a Study Mode feature aimed at discouraging cheating and promoting genuine learning. The update comes years after the chatbot reached 900 million weekly users with no teen-specific protections, amid lawsuits linking AI chatbots to youth mental health crises and suicides. The new product includes homework reminders, parental controls, learning visualizations, and age-appropriate content filters based on OpenAI's Under-18 Principles. However, questions remain about how easily teens can bypass these systems, given their track record of circumventing digital restrictions.

NEW AI PLATFORM LAUNCH

Cursor Launches GitHub Rival Amid Major Outage

Cursor rolled out Origin, its own code hosting platform, to paid users Monday morning. Hours later, GitHub suffered a nearly seven-hour global outage with error rates hitting 20% on core features and 50% on file downloads, taking down Copilot and enterprise authentication tools with it. The timing sparked immediate buzz online, with Vercel CEO Guillermo Rauch joking that Origin, unlike GitHub, was actually online. Cursor team member Matt Palmer quipped about the fortuitous launch timing, as developers scrambled to cope with the widespread GitHub disruption.

AI DATA ACQUISITION

Google Pays $10M for Spirit Airlines' Internal Data

Google has won a bankruptcy auction to acquire vast amounts of internal business data from Spirit Airlines for $10 million, outbidding competitors including AI training firm Mercor. The haul includes 100 million company emails, 500 million Microsoft Teams chats, 30 million lines of code, and operational records. Google confirmed the data will be scrubbed of personally identifiable information before use and stressed it did not acquire passenger or loyalty program records. The purchase reflects a growing market for proprietary business data as AI companies race to find fresh training material beyond freely available online content.

AI CLIMATE SUSTAINABILITY

Google Uses AI to Eliminate Climate-Warming Contrails

Google has partnered with the UK government and aviation industry on Operation Blue Skies, a 30-month initiative using AI to reduce contrail formation over the North Atlantic. The program will use satellite imagery, flight data, and weather forecasts to identify contrail-prone atmospheric zones, allowing air traffic controllers to reroute flights accordingly. Focusing on the Shanwick airspace, which accounts for 5% of global contrail warming, the trial expects around 10,000 flights with 1-5% being redirected. The £5 million project involves NATS, the Met Office, Contrails.org, and researchers from Imperial College London and Cambridge University.

AI CLOUD PARTNERSHIP DEAL

Ryanair Signs Five-Year AI Deal With Google

Ryanair has agreed to a five-year cloud and AI partnership with Google, giving its 35,000 employees access to Google Workspace, Google Cloud, and Gemini Enterprise. The AI platform will automate decision-making, optimize flight crew logistics, and support corporate productivity. Google DeepMind models will also assist with fleet operations and maintenance scheduling. The deal supports Ryanair's ambition to grow annual passenger numbers to 300 million by 2034. CEO Eddie Wilson said the partnership forms part of a dual-cloud strategy focused on infrastructure resilience and efficiency. Financial terms were not disclosed.

GENERATIVE AI WATERMARKING

Anthropic reveals Claude's invisible AI text watermarks

Anthropic has clarified its plan to embed invisible watermarks in Claude-generated text to comply with Europe's AI transparency regulations. The company announced the system on Friday. The watermarking approach is based on Google DeepMind's open-source SynthID-Text technology, which creates detectable patterns by manipulating word probability distributions during text generation.

NEW GENERATIVE AI FEATURE

Google lets users remove AI watermarks now

Google announced users can now remove visible watermarks from AI-generated images, videos, and songs. The toggle, available for Nano Banana, Omni, and Lyria models, will work in Gemini and video editor Flow, with Search support coming soon. Users can access it via Settings > Media Watermark. Invisible SynthID watermarks and C2PA metadata remain intact, preserving transparency tools for identifying AI content. Google is also open sourcing a library called Credentio for developers to embed local validation in their apps. The move follows Anthropic's decision to watermark Claude-generated content for EU compliance.

Upcoming Data & AI Conferences

NVIDIA GTC Berlin 2026

Tue, Oct 20 - Thu, Oct 22, 2026 | Berlin, Germany

Curated with AI by EMC2 AI

Brought to you by Estevan McCalley

Share this issue

Follow Estevan

Like this? Subscribe for the weekly AI roundup.