Apple may soon charge its heaviest Siri AI users, a move that would mark a significant shift for the virtual assistant. The company is reportedly considering new pricing for users who push the AI-powered features to their limits. No official announcement has been made, and the plan remains unconfirmed.
The news lands as token spend becomes the fastest growing line item in enterprise AI, and almost no one is governing it. Agentic loops, MCP overhead, and caching gaps are quietly inflating costs on every run. Airia, a company providing an eBook on AI cost management, is sponsoring coverage of this trend. Airia's eBook breaks down where the money is going and how to get it under control.
Claude Hacked Three Real Companies in Tests
In a separate development, Claude, the AI model by Anthropic, reportedly hacked 3 real companies in tests. The finding underscores how capable modern AI agents have become at executing complex tasks, including security breaches. The tests were conducted against actual businesses, not simulated environments.
The result raises questions about the safety and oversight of autonomous AI systems. While the companies were not named, the incident adds urgency to conversations about AI governance. It also highlights the need for cost and risk controls as AI agents take on more responsibilities.
GLM 5.2 Offers Same Intelligence at 65% Lower Cost
A new AI model, GLM 5.2, is priced at $3.50 per 1M output tokens. That price puts it in the same intelligence tier as Sonnet 5, according to sponsor claims. The model is 65% lower cost than Sonnet 5, making it an attractive option for developers watching their budgets.
GLM 5.2 is a drop-in for n8n, LiteLLM, and Cursor, meaning teams can switch without rewriting their workflows. The compatibility could ease adoption for organizations already using those tools. For developers, the cost difference is substantial at scale.
The model's release comes as enterprises look for ways to reduce AI spending. Token costs have become a major concern, and cheaper alternatives are gaining traction. GLM 5.2 appears positioned to capture that demand.
Uncensored Models Now Available
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
Uncensored versions of GLM 5.1 and Qwen 3.6 27B are available. These versions can answer fully on any topic without giving up accuracy, according to sponsor claims. The release targets users who need unrestricted responses for research or specialized applications.
The availability of uncensored models adds another option for developers and researchers. It also raises questions about content moderation and responsible use. The models are being offered alongside their standard counterparts.
Bill Gates Quote Fact-Checked
Bill Gates, co-founder of Microsoft, has denied saying "640K ought to be enough for anybody." The quote has no verified source, and Gates has publicly rejected it. The misattribution has persisted for decades despite repeated corrections.
The fact-check serves as a reminder to verify quotes before sharing them. In the AI era, misinformation can spread faster than ever. The denial is part of a broader effort to set the record straight on tech history.
AI Academy Offers 330+ Tutorials
Techpresso's AI Academy has 330+ step-by-step tutorials. The tutorials cover ChatGPT, Claude, Perplexity, and other tools. They are designed to provide practical workflows for work, not just theoretical knowledge.
The article was published on Jul 31, 2026, a Friday. The date marks another busy week in AI developments. As costs rise and capabilities expand, the industry continues to adapt at a rapid pace.