AI Newsletter

Goldman Sachs reverses its AI coding policy and restricts tools like Claude and GitHub Copilot to engineers – traders and bankers lose access, despite the bank publicly praising it as a productivity boost in July. The policy reversal suggests unrecognized risks (security, compliance, error rates) that emerged from uncontrolled AI usage in sensitive financial processes.

Musk throttles Grok Bot after just two weeks despite the product posting growth records – the reason is prohibitively high operating costs relative to revenue. This reveals an industry-wide critical problem: monetization of AI applications lags far behind enormous infrastructure costs, as OpenAI and Anthropic also struggle with.

Anthropic records impressive revenue growth (12x) but operates at massive loss ($23B), typical for AI startups in scaling phase. The upcoming IPO comes under critical scrutiny with these numbers, as investors demand paths to profitability. The high losses underscore the enormous capital requirements in the AI arms race between Anthropic, OpenAI, and other players.

The experiment reveals fundamental limitations of current AI agents in complex coordination tasks – they fail at self-organized group scenarios, while performing aggressively in direct competition. This exposes a critical weakness for enterprise deployments, where multi-agent systems are needed to automate complex workflows.

After US midterm elections, OpenAI launches an intensified lobbying and PR campaign to prevent or influence stricter AI regulation. This points to growing political risks that could threaten its business model and product development. The use of lobbying signals that regulatory hurdles are becoming an existential threat to OpenAI.