96% of Developers Don't Trust AI Code, CIOs Urged to Set Rules

Key Takeaways

• <ul><li>96% of developers do not fully trust AI-generated code</li><li>Only 48% always check AI code before using it</li><li>Synchrony pilots show 25% faster delivery and 30% fewer defects with AI</li><li>47% of physicians say their AI policies are still evolving</li><li>85% of people in China see more benefits than risks in AI vs. 38% in the US</li><li>Ambience Healthcare's scribe was used in over 3.5 million patient encounters in its first year</li><li>OpenAI's Chris Lehane pushes AI agenda ahead of 2028 election</li><li>New AISES-T scale measures teacher confidence with AI across three areas</li><li>AI-driven retail consolidation targets apparel, footwear, jewelry and wearables brands</li><li>David Donoho identifies five ways AI can boost human abilities</li></ul>

CIOs must guide AI coding to keep software quality high

AI is writing more code, but 96% of developers do not fully trust AI-generated code. Only 48% always check AI code before using it. Experts say CIOs need clear rules, training, and careful oversight. Experienced engineers are still needed to review and improve AI-written code.

AI will not end developer jobs, Synchrony leader says

AI will not replace developers but will change their role into software engineers. At Synchrony, pilots showed 25% faster delivery, 30% fewer defects, and 40% fewer security problems. Humans must still test AI output and set guardrails. The article calls for trust, mentorship, and teaching AI skills from the ground up.

Public views on AI differ sharply across countries

A 2026 Ipsos survey found 85% of people in China see more benefits than risks in AI, compared with only 38% in the US. The UK, France, and the Netherlands also showed low AI optimism. The article says attitudes depend on what a society already has, not just income or growth. AI can replace costly experts in rich nations but supply new expertise where it was missing.

Military AI can lead troops to wrong targets, adviser warns

Andrew Moore, former AI adviser to US CENTCOM, warns that asking AI the wrong question can produce misleading results. AI could return thousands of pieces of evidence supporting one idea, even if the target is wrong. He says AI still needs strong testing, like other military tools. Speed helps, but it does not guarantee correct decisions.

New scale measures teacher confidence in using AI

A 2026 study in Frontiers in Education introduced AISES-T, a 22-item scale to measure teacher confidence with AI. It splits confidence into three areas: using AI tools, integrating AI into lessons, and understanding AI concepts. The tool helps schools see where teachers need training. It shows confidence is different from skill or attitude.

AI boom drives retail consolidation as middle brands face acquisition

AI-driven wealth is splitting the consumer economy and creating acquisition targets among struggling brands. Strong operators and private equity buyers are looking to roll up brands with awareness but lacking capital. Apparel, footwear, jewelry and wearables look most exposed. Higher rates from the Federal Reserve's September 16 decision make debt-heavy deals harder. The next 12 to 18 months could see strong brands grow while weaker ones disappear or get absorbed.

CNN asks readers to share how AI agents shape their daily routines

Tech giants believe AI agents will play a major role in how people handle daily tasks like shopping, returning packages and planning vacations. Virtual assistants such as Siri and Alexa are already widely used. CNN is asking readers to share their personal experiences using AI agents in everyday life. CNN's Brian Fung contributed to the report.

States create new rules as AI scribes listen in on doctor visits

AI scribes now record doctor-patient conversations and draft medical notes. Louisiana requires spoken disclosure before recording, Rhode Island requires patient notice, and California requires disclaimers on unread AI messages. Cleveland Clinic used Ambience Healthcare's scribe in over 3.5 million patient encounters in its first year. Nearly 47 percent of physicians say their AI policies are still evolving. HIPAA covers patient data but does not require specific AI scribe notifications.

OpenAI's Chris Lehane pushes AI agenda ahead of 2028 election

OpenAI's chief global affairs officer Chris Lehane says the 2028 election could define AI's role in America. He proposes tax incentives for workers to learn AI skills and supports expanding data centers. Critics like the AI Now Institute's Kate Brennan say his ideas favor OpenAI's interests and lack accountability measures. Affordability and data center costs are expected to be key 2028 issues. Lehane also calls for child safety and education measures around AI.

Mathematician David Donoho sees hope for human intellect in AI era

Vanderbilt math chair David Donoho lists five ways AI can boost human abilities rather than replace them. He points to improvements in data analysis, pattern recognition, mathematical modeling, human-AI collaboration and education. Donoho believes AI is a tool that helps people reach their full potential. He has studied AI's impact on human intellect for over a decade.

Sources

• How CIOs can retain control as AI writes more code
• AI won’t end developer jobs
• Mapping AI resistance
• Former CENTCOM AI Adviser Warns Troops Asking AI the Wrong Question Could Point Them at the Wrong Target
• AISES-T: 22-Item Teacher AI Confidence Scale [2026]
• AI Is Fueling Retail’s Next Consolidation Wave
• Do you use AI agents to run your daily life? We want to hear about it
• As AI Listens In on Doctor Visits, States Are Writing the Rules
• OpenAI is already trying to shape the 2028 election
• A mathematician finds hope after AI's new dawn