After reading this, the reader knows Spain is reviewing an AI-agent-linked breach, and it matters because one agent allegedly chained several attack stages. Spain reports AI-agent-linked data breach The notification says an agent logged in, searched for vulnerabilities, changed personal data and viewed invoices. The model and victim remain unidentified. Spain’s data protection authority has received what it describes as the country’s first reported notification of a personal-data breach allegedly carried out with an AI agent, according to the authority’s account relayed by Reuters. ...

After reading this, the reader knows Apple added verified photo capture, and it matters because synthetic images have weakened visual evidence. Apple launches verified photo capture mode Reference Image signs sensor data, timestamps capture and processes the result in Private Cloud Compute without publicly identifying the photographer. Apple announced Reference Image on September 15, an opt-in camera mode for the iPhone 18 Pro and Pro Max designed to prove that a photograph originated from a physical sensor at a bounded time. ...

After reading this, the reader knows Cloudflare separated search and training controls, and it matters because publishers can reject training without disappearing from search. Cloudflare separates search from AI training The new setting distinguishes search, model training and user-directed agents. Support still depends on crawler operators honoring declared preferences. Cloudflare introduced a Disallow AI Training setting on September 15 that lets website owners keep compatible search crawlers while expressing a refusal to let the same operators use their content for model training. ...

After reading this, the reader knows Salesforce built Koa for CRM tool use, and it matters because enterprise agents need specialized action models. Salesforce launches Koa for CRM agents The model starts from NVIDIA Nemotron 3 Super and is post-trained on synthetic workflows. Salesforce says no customer data entered its training set. Salesforce and NVIDIA announced Koa on September 15, a specialized language model intended to reason through multi-step customer-relationship-management tasks and call software tools inside Agentforce. ...

After reading this, the reader knows Strix says it found a live GitHub token in a public Baseten container image, and how build history can expose secrets. Strix finds GitHub token in Baseten image The security company says it stopped after read-only checks, disclosed the issue and saw Baseten rotate the credential and restrict its registry the next day. Security company Strix says its autonomous testing agent found a live GitHub personal access token in the build history of a publicly downloadable Baseten container image. ...

After reading this, the reader knows growing inference demand is pushing chip designers toward faster memory movement and specialized prefill and decode hardware. AI inference pushes memory-centric chips An IEEE Spectrum survey shows vendors stacking memory with compute, using on-chip SRAM and splitting language-model serving across different processors. AI hardware designers are reorganizing systems around memory bandwidth as inference becomes a larger share of data-center work, according to an IEEE Spectrum analysis published September 15. ...

After reading this, the reader knows an agent-world experiment reportedly produced repeated model-specific phrases, raising questions about monitoring persistent autonomous systems. Agent societies develop opaque shorthand Emergence AI says agents formed recurring phrases and conventions without explicit instruction. The specific language findings await a primary technical release. Autonomous AI agents developed repeated phrases and shared conventions in a persistent simulation, according to an Emergence AI experiment reported by The Guardian on September 15. ...

After reading this, the reader knows OpenAI, Anthropic and Google are reportedly discussing AI safety, while the scope and outcome remain unconfirmed. AI labs reportedly open joint safety talks A report says three leading model developers have held discussions for weeks. None had publicly confirmed a concrete agreement when this article was prepared. OpenAI is reportedly in talks with Anthropic and Google about cooperation on AI safety, according to a Bloomberg report relayed by Reuters on September 15. ...

After reading this, the reader knows Delos Data reportedly raised $100 million to move data faster inside AI systems, and why memory traffic is becoming a chip priority. Delos Data raises $100M for AI networking The Intel-veteran startup is building chips and software for data movement among accelerators, memory and storage in AI data centers. Delos Data said it raised $100 million to develop networking technology for AI data centers, according to Reuters on September 15. ...