OpenAI expands workflow tools as Anthropic highlights security concerns

OpenAI introduces GPT-6.1 Terra as an assistant designed to coordinate computer-based tasks across browsers, spreadsheets and document editors. GPT-Live receives a voice update focused on pronunciation and more natural conversation, while ChatGPT adjusts its tone to reduce unnecessary flattery and deliver more direct answers. OpenAI also updates its model specifications governing assistant behavior, including safety, honesty, privacy and responses to user instructions. Availability and support vary across products, models and plans.

OpenAI moves its Decisions API beyond public beta to general availability, enabling developers to build applications that evaluate actions and route tasks through workflows. Eligible subscribers may receive up to $500 in API credits, subject to eligibility, expiration and usage restrictions. Partnerships involving Oracle and ServiceNow aim to connect conversational AI with enterprise processes across finance, customer service, IT and operations. Deployment depends on reliable integrations, permissions, privacy protections and governance controls.

Anthropic discloses a training pause connected to concerns about security safeguards and testing environments, underscoring the importance of controlled access and infrastructure security. Cloud marketplaces accounted for a reported 48% of first-half 2026 revenue, highlighting their importance as distribution channels without implying company ownership. Meanwhile, investors are scrutinizing spending on data centers, chips, energy and supporting infrastructure, questioning whether expected returns can justify expansion costs.