Google's Threat Intelligence Group is sounding the alarm on how AI is reshaping cybersecurity. According to their recent findings, vulnerabilities discovered with the help of AI are more likely to result in remote code execution than those found through traditional methods. The group reported that monthly CVE disclosures nearly doubled, jumping from 5,045 in January 2026 to 10,740 by August 2026. Threat actors have been quick to exploit these AI-discovered flaws, sometimes within just four days of disclosure.
Google's report also highlights how AI helps cybercriminals exploit known software bugs, so-called n-days, more quickly. Zero-day exploitation grew from 8 per month in 2025 to 11 per month in 2026. AI-discovered flaws led to remote code execution 50% of the time, compared to 26% overall. Google now recommends threat-intelligence-driven triage instead of mass patching to address the problem.
In other AI safety news, technology leaders including Google CEO Sundar Pichai and Microsoft CEO Satya Nadella signed a voluntary AI safety agreement. The non-binding deal aims to increase transparency and accountability in AI development. However, some experts say the accord does not go far enough, especially as AI agents gain popularity for automating tasks and raise concerns about job displacement and misuse.
On the product side, NVIDIA released DOCA Agent Skills to help developers build applications on NVIDIA BlueField DPUs more accurately and safely. In a 65-prompt evaluation, the skills prevented AI agents from inventing fake API functions, writing code for unsupported hardware features, or using wrong build settings. The skills also guide agents through safe firmware changes on live hardware by requiring preflight checks and rollback plans.
Real Chemistry launched MaestroHCP, an AI-powered platform designed to streamline omnichannel marketing campaigns. The tool aims to replace the manual and fragmented process of running campaigns across social media, email, and websites. Real Chemistry CEO Stan Woodland said the platform will change how marketers approach these campaigns and marks a major step for the company in using AI to improve marketing efficiency.
Separately, Palo Alto Networks introduced Cortex XCOR, an AI-driven observability platform that automatically finds and fixes problems in cloud systems. The platform features an AI assistant called XCOR Operator that works with specialized agents to handle tasks like investigating incidents and tuning alerts. The company says the platform delivers the long-held goal of automatic root cause analysis, a need that has grown as more developers use AI to write code.
McKinsey senior partner David Pralong warned that CEOs are mistaking easy AI successes for a universal playbook. He noted that AI works well in contact centers and coding because those jobs already had clear queues and review steps. But faster task completion does not fix broken processes that cross many departments. Pralong said leaders must redesign work, assign ownership, set measures, and explain what changes for employees before AI pilots can create real value.
In healthcare, Dr. Benjamin Hansen at UMC in Las Vegas uses AI-powered OrthoGrid technology to assist with hip replacement surgeries. The tool automates measurements that were once done manually, improving accuracy in leg length and offset. More than 250 physicians in the country have used OrthoGrid across 100,000 procedures. However, Dr. Hansen says the technology has limits and cannot replace a surgeon's training and experience. High costs also create a divide between wealthy hospitals and rural clinics.
New government AI policy is creating opportunities for training providers. The federal government has directed agencies to call AI super intelligence in official documents, while states have passed 85 new AI laws across 27 states this year. States like Illinois, Texas, and California have enacted enforcement rules that create real compliance needs for businesses. The European Commission is also gathering feedback on its AI Act, adding more complexity for global organizations.
On the robotics front, a Chinese AI model called Maxwell, developed by the Chinese Academy of Sciences, topped a global ranking for physical tasks. Maxwell outperformed other models in grasping and manipulation tasks. Researchers say its success comes from learning from experience and adapting to new situations. The study was published in the journal Science, and the findings could advance robots and machines that perform complex physical tasks.
Finally, a real estate expert warns that AI-generated property media can hurt buyer trust when accuracy suffers. AI-generated videos sometimes add objects or change rooms in ways buyers can spot. The author recommends keeping original photos and having a broker review listings before publishing. Accurate descriptions matter more than polished but misleading media, the expert said.
Key Takeaways
- Google's Threat Intelligence Group found that AI-discovered vulnerabilities lead to remote code execution 50% of the time, compared to 26% overall.
- Monthly CVE disclosures nearly doubled from 5,045 in January 2026 to 10,740 in August 2026, according to Google.
- Google CEO Sundar Pichai and Microsoft CEO Satya Nadella signed a voluntary AI safety agreement aimed at increasing transparency and accountability.
- NVIDIA's DOCA Agent Skills prevented AI agents from inventing fake API functions in a 65-prompt evaluation, improving safety for BlueField DPU development.
- Real Chemistry launched MaestroHCP, an AI-powered platform designed to streamline omnichannel marketing campaigns across social media, email, and websites.
- Palo Alto Networks introduced Cortex XCOR, an AI-driven observability platform featuring an AI assistant called XCOR Operator for investigating cloud incidents.
- McKinsey partner David Pralong warns that faster AI task completion does not fix broken cross-departmental processes without broader work redesign.
- More than 250 physicians have used OrthoGrid across 100,000 procedures, but the AI tool cannot replace a surgeon's training and experience.
- States have passed 85 new AI laws across 27 states this year, creating compliance needs that training providers can help businesses address.
- Chinese AI model Maxwell topped a global ranking for physical tasks like grasping and manipulation, with findings published in the journal Science.
AI-found vulnerabilities are the ones attackers want, Google says
Google Threat Intelligence Group found that AI-discovered vulnerabilities are more likely to lead to remote code execution than flaws found by other methods. Monthly CVE disclosures doubled from 5,045 in January 2026 to 10,740 in August 2026. Threat actors exploited one AI-found flaw within four days of disclosure. Attackers used tools like SNOWLIGHT and SPARKRAT in those campaigns. AI orchestration tools such as vLLM and Ollama are now the top target for vulnerabilities.
Google warns AI helps attackers weaponize known bugs faster
Google's GTIG report says AI helps cybercriminals exploit known software flaws, called n-days, more quickly. The number of bugs found monthly rose from 5,045 in January to 10,740 in August 2026. Zero-day exploitation grew from 8 per month in 2025 to 11 per month in 2026. AI-discovered flaws led to remote code execution 50% of the time, compared to 26% overall. Google recommends threat-intelligence-driven triage instead of mass patching.
AI can speed up real estate listings but buyers notice errors
A real estate expert warns that AI-generated property media can hurt buyer trust when accuracy suffers. Original photographs should serve as the reference for any AI-enhanced videos or images. AI-generated videos sometimes add objects or change rooms in ways buyers can spot. The author recommends keeping original photos and having a broker review listings before publishing. Accurate descriptions matter more than polished but misleading media.
Chinese AI model Maxwell tops global ranking for physical tasks
A new Chinese AI model called Maxwell, developed by the Chinese Academy of Sciences, topped a global ranking for physical tasks. Maxwell outperformed other models in grasping and manipulation tasks. Researchers say its success comes from learning from experience and adapting to new situations. The study was published in the journal Science. The findings could advance robots and machines that perform complex physical tasks.
Real Chemistry launches AI platform MaestroHCP for marketing
Real Chemistry launched MaestroHCP, an AI-powered platform for omnichannel marketing campaigns. The platform aims to streamline the manual and fragmented process of running campaigns across channels like social media, email, and websites. Real Chemistry CEO Stan Woodland said the tool will change how marketers approach omnichannel campaigns. The launch marks a major step for the company in using AI to improve marketing efficiency.
Palo Alto Networks launches AI-native observability platform Cortex XCOR
Palo Alto Networks introduced Cortex XCOR, an AI-driven observability platform that automatically finds and fixes problems in cloud systems. The platform features an AI assistant called XCOR Operator that works with specialized agents to handle tasks like investigating incidents and tuning alerts. This launch comes as more developers use AI to write code, increasing the need for smarter monitoring tools. The company says the platform finally delivers the long-held goal of automatic root cause analysis.
NVIDIA DOCA Agent Skills help developers build BlueField applications faster
NVIDIA released DOCA Agent Skills to help developers build applications on NVIDIA BlueField DPUs more accurately and safely. In a 65-prompt evaluation, the skills prevented AI agents from inventing fake API functions, writing code for unsupported hardware features, or using wrong build settings. The skills also guide agents through safe firmware changes on live hardware by requiring preflight checks and rollback plans. These improvements help developers move from debugging to integration more quickly.
Tech leaders sign voluntary AI safety accord as AI agents gain popularity
Technology executives including Google CEO Sundar Pichai and Microsoft CEO Satya Nadella signed a voluntary AI safety agreement outlining principles for AI development. The non-binding deal aims to increase transparency and accountability, though some experts say it does not go far enough. Meanwhile, AI agents are growing in popularity for automating tasks, raising concerns about job displacement and misuse. Experts also warn that AI could affect politics as extremist candidates run in the US midterms.
New government AI policy creates opportunities for training providers
The federal government has directed agencies to call AI super intelligence in official documents, while states have passed 85 new AI laws across 27 states this year. States like Illinois, Texas, and California have enacted enforcement rules that create real compliance needs for businesses. Training providers can offer certification courses to help companies meet these requirements. The European Commission is also gathering feedback on its AI Act, adding more complexity for global organizations. This gap between federal language and state rules creates a market for practical training.
AI helps Las Vegas orthopedic surgeon but cannot replace training
Dr. Benjamin Hansen at UMC in Las Vegas uses AI-powered OrthoGrid technology to assist with hip replacement surgeries. The tool automates measurements that were once done manually, improving accuracy in leg length and offset. However, Dr. Hansen says the technology has limits and cannot replace a surgeon's training and experience. More than 250 physicians in the country have used OrthoGrid across 100,000 procedures. High costs also create a divide between wealthy hospitals and rural clinics.
McKinsey partner warns AI quick wins mislead CEOs
David Pralong, a senior partner at McKinsey & Company, warns that CEOs are mistaking easy AI successes for a universal playbook. AI works well in contact centers and coding because those jobs already had clear queues and review steps. But faster task completion does not fix broken processes that cross many departments. He gives a bank example where faster document review still left customers waiting through handoffs among sales, credit, compliance, and operations. Leaders must redesign work, assign ownership, set measures, and explain what changes for employees before AI pilots can create real value.
Sources
- The vulnerabilities AI finds are the ones attackers want
- 'It is possible that threat actors are finding it more accessible or efficient to use LLMs and AI tools': Google warns that AI explosion will lead to more dangerous and advanced security threats
- AI can speed up listing media, but buyers notice when it is off
- New Chinese embodied model Maxwell tops global AI ranking on physical tasks
- Real Chemistry launches AI-powered MaestroHCP to conduct omnichannel campaigns
- Observability’s AI Moment
- Build Applications on NVIDIA BlueField Faster with NVIDIA DOCA Agent Skills
- Trump’s ‘Morally Binding’ AI ‘Accord,’ the Rise of AI Agents, and Extremists on the Ballot
- New Government AI Policy Just Dropped! What It Means for Training Providers Selling Compliance Adjacent Certification
- AI assists surgeons at UMC, but Las Vegas doctor explains why it can’t replace training
- McKinsey senior partner: Why AI’s easiest wins are misleading CEOs
Comments
Please log in to post a comment.