Google Gemini security flaws while Hugging Face enhances safety

Key Takeaways

• 46% of AI-generated code contains security flaws, prompting solutions like Harness's platform, which scans code and supports tools like Google's Gemini and GitHub Copilot.
• Apiiro's new CLI and Appknox's KnoxIQ integrate AI-native security directly into development workflows, offering real-time vulnerability scanning, risk assessment, and automated code fixes.
• The future of AI in sales requires specialized models for specific industries, moving beyond general LLM wrappers, as noted by Google VP Darren Mowry.
• Salesforce is evolving into an AI agent platform with Agentforce, leading an "agentic CRM revolution" alongside new CRMs like Aurasell and Monaco.
• AI is crucial for growth in the e-commerce resale market, with companies like ThredUp and The RealReal using it for automation, authentication, pricing, and customer experience.
• AI is reshaping the cybersecurity job market, automating entry-level tasks and increasing demand for specialized leadership roles in AI and cloud computing.
• Hugging Face's Safetensors project has joined the PyTorch Foundation to enhance AI model execution security, preventing arbitrary code execution and improving performance.
• The emerging "Bring Your Own AI" (BYOAI) trend suggests personal AI tools may migrate to the workplace, driven by convenience but raising security and privacy concerns.
• Financial institutions are adopting unified, AI-powered platforms, with insights from experts like AWS, to manage fraud, cybersecurity, and compliance in real-time.
• Wharton Online offers an AI and Digital Strategy Certificate, and Manulife Investment Management is piloting AI for ESG assessments, highlighting the strategic integration of AI across sectors.

AI code 46% vulnerable, Harness offers security solutions

A recent study found that 46% of code generated by AI tools contains security flaws. Jignesh Patel from Harness explained that securing this code requires integrating security throughout the entire development process, not just scanning it. Harness's platform scans code as it's created and integrates with other security tools. They support various AI platforms like Gemini and GitHub Copilot, aiming to make AI-assisted development safer.

Apiiro CLI brings AI security into development workflows

Apiiro has launched a new command-line interface (CLI) to integrate AI-native security directly into software development. This tool helps AI agents make smarter security decisions as they write code, addressing the gap left by traditional security tools. The Apiiro CLI allows AI assistants to access security information, apply policies, and fix issues in real time. It offers features like real-time vulnerability scanning, risk assessment, and automated code fixes, aiming to embed security from the start.

Appknox launches KnoxIQ for AI-driven application security

Appknox has released KnoxIQ, a new tool that uses AI to prioritize application security vulnerabilities based on real-world risks. As AI generates more code, it also creates more complex security issues. KnoxIQ helps security teams by reducing false positives and providing ready-to-use code fixes. It integrates with AI developer tools and analyzes applications based on their runtime behavior for better accuracy. This approach aims to make security more efficient and actionable for developers.

AI sales future is diverse and distributed, not just LLM wrappers

The future of AI in sales will rely on a diverse range of AI technologies, not just simple 'LLM wrappers' or aggregators. Google VP Darren Mowry warned that startups relying solely on existing large language models are at risk. The most successful AI companies will build specialized models for specific industries. This shift emphasizes using the right AI tool for each task, moving beyond general-purpose models to more focused and effective solutions.

Agentic CRM revolution at SaaStr AI Annual 2026

The CRM market is being rebuilt with AI agents, and leaders will gather at SaaStr AI Annual 2026 from May 12-14. Salesforce is evolving into an AI agent platform with Agentforce, integrating acquired tools for better context and data sharing. Next-gen CRMs like Aurasell and Monaco are also emerging. Aurasell aims to replace multiple GTM tools with one AI-native platform, while Monaco combines AI software with human salespeople for supervision. The event will showcase companies leading this agentic CRM revolution.

AI fuels growth in ecommerce resale market

Artificial intelligence is playing a key role in the expanding resale market within e-commerce. Companies like ThredUp and The RealReal are using AI to automate tasks and improve customer experience. ThredUp found that many shoppers are comfortable with AI managing their resale activities, prioritizing speed and ease. The RealReal uses AI for authentication, pricing, and sales lead scoring, aiming to increase efficiency and personalize the shopping experience. AI is seen as crucial for the next phase of online resale growth.

AI impacts cybersecurity jobs, creating new roles and risks

Artificial intelligence is significantly changing the cybersecurity job market. While AI strengthens defenses, it also reduces entry-level positions by automating tasks like analysis and triage. This creates a skills gap and makes it harder for new professionals to enter the field. Highly compensated leadership roles are increasing, requiring specialized skills in areas like AI and cloud computing. Cybersecurity hiring continues to lag behind demand, with skills becoming obsolete quickly, leading to a focus on specialized talent over traditional job titles.

Safetensors joins PyTorch Foundation for secure AI model execution

The PyTorch Foundation has welcomed Safetensors as its newest project to enhance the security of AI model execution. Developed by Hugging Face, Safetensors prevents arbitrary code execution risks that were present in older file formats. This new format acts like a table of contents for AI model data, making distribution safer and improving performance across multiple GPUs. Safetensors is becoming a standard for open-weight model distribution, ensuring faster and more secure AI development.

BYOAI: Will personal AI move to the workplace?

The concept of 'Bring Your Own AI' (BYOAI) is emerging, similar to the 'Bring Your Own Device' (BYOD) trend. As people become accustomed to using personal AI tools with their own data and workflows, they may want to use these same tools at work. While security and privacy are concerns, the convenience of familiar AI could drive adoption, especially for smaller businesses. This trend could influence how AI is used in the workplace and potentially shift AI usage from work to home.

Manulife IM uses AI for ESG assessments

Manulife Investment Management is continuing to use proxy advisors despite exploring AI for proposal pushes. The company's Chief Sustainability Officer (CSO) discussed a pilot program using AI to integrate ESG materiality assessments. This suggests a growing interest in leveraging AI for more sophisticated analysis within investment management, even while relying on established advisory methods.

Wharton launches AI and Digital Strategy Certificate

Wharton Online has launched a new certificate program focused on Digital Strategy in the Era of AI. This four-course, self-paced program helps professionals lead digital transformation and make better decisions using AI. It emphasizes strategic thinking, organizational alignment, and practical execution, rather than just tools and trends. The certificate is designed for leaders responsible for strategy and change, requiring no technical background, and aims to equip participants to navigate the complexities of AI-driven business environments.

AI security needs governance and awareness, says CISO

Deepesh Randeri, CISO at Akron Children's Hospital, emphasizes that AI security requires strong governance and awareness from the start. He explains that all new technologies, including AI, must undergo a rigorous vetting process to ensure security standards are met. Randeri highlights the importance of monitoring AI systems throughout their lifecycle, not just during implementation, to prevent bias and ensure guardrails are in place. He stresses that a strong 'tone at the top' is crucial for embedding security into AI adoption.

AI unifies banking fraud, security, and compliance

Financial institutions are using unified, AI-powered platforms to manage the increasing overlap between fraud, cybersecurity, and compliance. Experts from IDC, Elastic, and AWS discussed how banks are building real-time, AI-driven capabilities to detect and stop fraud. The session covered practical steps for aligning fraud, security, and compliance teams, how AI is reshaping risk management, and how banks can stay ahead of evolving regulations. This approach aims to build greater resilience and trust in banking.

Sources

• Is 46% of your AI-generated code vulnerable?
• Apiiro launches Apiiro CLI to bring AI-native security into software development workflows
• Appknox launches KnoxIQ to prioritize real-world exploitability in AI-driven application security
• The Future of AI for Sales Is Diverse and Distributed
• Meet the Leaders of the Agentic CRM Revolution at SaaStr AI Annual 2026, May 12-14 in SF Bay!
• Ecommerce Trends: What is AI's role in resale growth?
• AI’s double edged impact on cybersecurity jobs
• PyTorch Foundation Announces Safetensors as Newest Contributed Project to Secure AI Model Execution
• BYOAI
• Manulife IM sticking with proxy advisers despite AI proposal push
• Wharton Online Launches Digital Strategy in the Era of AI Certificate
• AI security starts with awareness and governance, CISO says
• Intelligent Banking in the Age of AI: Unifying Fraud, Security, and Compliance