🧠 Food for thought

Implications, context, and why it matters.

Chinese AI stocks rally on price and performance, not hype

• Fresh model launches plus low-price claims are driving the run-up, with a direct challenge to Western leaders.
• Zhipu AI shares rose more than 100% in a week after it released a new model 1.
• MiniMax M2.5 reached 80.2% on SWE-Bench Verified (a benchmark for evaluating how well AI models solve real software engineering tasks), close to Anthropic Claude Opus 4.6 at 80.8% in third-party comparisons 2.
• Those comparisons also peg M2.5 at about 1/20th the cost using published pricing examples for finishing a SWE-Bench task 2.
• Demand has been strong enough that Zhipu raised prices for its coding plan by 30% 1.

Open-source AI agents add ‘shadow IT’ risk and slip past security controls

• Tools such as OpenClaw can weaken business IT safeguards when set up poorly or given broad system permissions.
• China’s Ministry of Industry and Information Technology’s National Vulnerability Database (NVDB) warned that default or improper OpenClaw setups can invite cyberattacks and data leaks under default or improper configurations 3. It urged tighter authentication, access control, encryption, auditing 3.
• Unapproved employee deployments can create “Shadow AI” exposure, which can widen access outside standard controls 4.
• Researchers warn that indirect prompt injection, such as via malicious emails, can steer an agent into unintended actions when it processes untrusted inputs with high privileges 5.
• Marketplace risk adds more openings. One analysis found 336 malicious samples in over 3,000 ClawHub Skills, about 10.8%, raising supply-chain poisoning concerns 5.