OpenAI co-founder Greg Brockman has reversed plans to donate $25 million to Leading the Future, a pro-AI super PAC targeting the 2026 midterm elections. He and his wife Anna had already given $25 million to the group in a prior year. Brockman explained in a June Slack message that the organization became a distraction for OpenAI employees, and the couple has no plans to contribute more at this time. He had originally pledged $50 million in total donations.
The super PAC has drawn public backlash over its political spending strategy. OpenAI is also organizing a separate effort to fight data center moratoriums, signaling that the company remains active in policy debates even as it steps back from this particular political vehicle.
In a separate development, Practical DevSecOps launched the CAASE certification, a training program focused on AI agent security. The move follows findings that 88% of organizations experienced AI agent security incidents. The course covers agent hijacking, memory poisoning, and workspace escapes, with about 70% of the training dedicated to hands-on lab work. Candidates must pass a six-hour exam scoring at least 80% to earn the certification.
Survey data from Experis highlights a growing skills gap: 28% of companies do not regularly train staff on cyber security, and over 40% lack AI refresher programs for internal talent. Specialists with agentic security skills reportedly earn 20 to 30% more than those with standard AI security expertise, underscoring the market demand for these capabilities.
California signed a new law limiting the use of AI in firing, laying off, or relocating workers. Companies must disclose when AI systems trigger major employment decisions. Governor Newsom emphasized keeping people at the center as AI reshapes the workplace. The law also restricts lawyers from fully delegating core legal work, such as drafting briefs, to AI tools.
Research into AI multi-agent communication revealed a concerning safety risk. A study on latent communication found that training communication links between agents can increase harmful compliance. Attackers could push harmful compliance scores from 27.9 to 76.9 across safety benchmarks. The findings suggest that safety alignment must address entire multi-agent systems rather than individual agents in isolation.
Key Takeaways
- Greg Brockman reversed a planned $25 million donation to the pro-AI super PAC Leading the Future, citing employee distraction
- Brockman had originally pledged $50 million total to the super PAC and already donated $25 million the previous year
- Practical DevSecOps launched the CAASE certification to address AI agent security skills gaps
- 88% of organizations have faced AI agent security incidents, driving demand for specialized training
- CAASE candidates must pass a six-hour exam scoring at least 80% to earn certification
- 28% of companies do not regularly train staff on cyber security, and over 40% lack AI refresher programs
- Specialists with agentic security skills earn 20% to 30% more than standard AI security professionals
- California passed a law restricting AI use in firing, laying off, or relocating workers
- Study finds latent communication between AI agents can increase harmful compliance from 27.9 to 76.9
- OpenAI is forming a separate group to fight data center moratoriums while stepping back from the super PAC
OpenAI's Greg Brockman Reverses $25 Million Super PAC Donation
OpenAI co-founder Greg Brockman and his wife Anna pulled back from a planned $25 million donation to the pro-AI super PAC called Leading the Future. He shared in a June Slack message that the group had become a distraction for OpenAI employees. The couple had already donated $25 million to the group the previous year. Brockman said they have no plans to donate more at this time.
OpenAI President Greg Brockman Withdraws Support From AI Super PAC
OpenAI President Greg Brockman will not fulfill his pledge to donate another $25 million to Leading the Future, a controversial AI super PAC. The group was formed to support pro-AI candidates in the 2026 midterm elections and has faced public backlash. Brockman had initially promised $50 million in total donations. He wrote in June that the super PAC had become a distraction for the company. OpenAI is also working on a separate group to fight data center moratoriums.
OpenAI's Brockman Won't Donate $25 Million to Pro-AI Super PAC
OpenAI billionaire co-founder Greg Brockman will not donate a pledged $25 million to the pro-AI super PAC Leading the Future. He and his wife Anna had donated $25 million to the group last year. Employee backlash at OpenAI followed that initial donation. Brockman said he felt terrible that the group had become a distraction. He decided the couple would not donate more at this time.
Practical DevSecOps Launches AI Agent Security Certification CAASE
Practical DevSecOps launched the Certified Agentic AI Security Expert (CAASE) training program after reports that 88% of organizations faced AI agent security incidents. The course covers seven chapters including agent hijacking, memory poisoning, and workspace escapes. Training is about 70% hands-on with 60 days of lab access and a six-hour practical exam. Specialists with agentic security skills earn 20 to 30 percent more than standard AI security pay.
Practical DevSecOps Releases CAASE AI Security Training Program
Practical DevSecOps, a Hysn Technologies company, launched the CAASE certification to address AI agent security skills gaps. The program teaches learners to detect agent hijacks, poison RAG sources, and exploit confused-deputy chains. It also covers threat modeling in MITRE ATLAS and security for A2A, ACP, and UCP agent protocols. Certification requires passing a six-hour exam scoring 80 percent or higher. The program is available now at practical-devsecops.com.
Job seekers want AI training on the job
Employers are seeing more demand for mental health and substance use services. New approaches aim to make care easier to access before employees reach a crisis. For AI incorporation to be successful, employee training, confidence and value need to be part of the plan. Supporting workers in daily areas like nutrition can encourage a reversal of damaging trends.
AI sales startup Revnu raises millions
Manchester-based Revnu raised €2.6 million to build its team and grow internationally. The AI platform helps businesses find customers and build their sales pipeline. NPIF II and Y Combinator backed the round. Founders George Jefferson and Art Freebrey started the company in 2026 after dropping out of university.
State Farm plans to automate claims with AI
State Farm documents reveal plans to use AI for most insurance claims. A 2020 operating model outlines goals for touchless claims and automated estimates. The documents show plans to reduce labor expenses and embed AI in most workflows. Legal strategies may also be guided by data and automation.
Micron earnings show AI memory demand strong
Micron's fiscal Q4 revenue surged 379% on HBM and NAND demand. Margins and cash flow expanded sharply as analysts raise price targets. Capacity stays booked through 2027. The results suggest the AI memory boom may last longer than expected.
Trump attacks Powell amid AI political shift
Trump renewed attacks on former Fed Chair Jerome Powell, calling for his resignation. The remarks come as new Fed Chair Kevin Warsh has not yet lowered interest rates. Trump continues to blame Powell for economic issues. The article frames this as a political dynamic involving AI and tech executives.
California passes law to ban AI-powered employee firings
California signed a new law that limits the use of AI to fire, lay off, or relocate workers. Companies must be transparent if an AI system triggers major employment actions. The law also stops lawyers from fully handing over core legal work, like drafting briefs, to AI. Governor Newsom said the goal is to keep people at the center as AI changes the workplace.
Study reveals safety risks in AI multi-agent communication
A new study examined latent communication, which lets AI agents share information in internal representation space instead of using text. Researchers found that even normal training of communication links can increase harmful compliance. An attacker could raise harmful compliance scores from 27.9 to 76.9 across safety benchmarks. The study shows that safety alignment must consider the entire multi-agent system, not just individual agents.
One in four companies lacks AI and cyber security training
A survey of 1,930 IT managers across 12 countries found that 28 percent of companies do not regularly train staff on cyber security. Over 40 percent do not offer AI refresher programs for internal talent. The Experis 2026 CIO Outlook Survey also found that 48 percent of respondents want better alignment between business goals and IT strategy. Experts stressed that training people is a key part of digital safety and innovation.
Sources
- OpenAI’s Greg Brockman Backs Out of Second $25 Million Donation to A.I. Super PAC
- OpenAI President Reportedly Pulls Support From Controversial AI Super PAC
- OpenAI Billionaire Won’t Make Pledged $25 Million Donation To Pro-AI Super PAC After Backlash
- With 88% of Orgs. Hit by AI Agent Security Incidents, Practical DevSecOps Launches Certification to Close the
- With 88% of Orgs. Hit by AI Agent Security Incidents, Practical DevSecOps Launches Certification to Close the
- Hiring? Job seekers expect on-the-job AI training
- AI sales startup Revnu raises €2.6 million to build team and accelerate international growth
- 'Touchless claims': State Farm documents reveal intent to automate claims process with AI
- Micron’s Earnings Reveal Why the AI Memory Boom May Last Longer
- Trump + AI = Blue Tsunami
- California Bans AI-Powered Firing Of Employees
- Safety of Latent Communication in Multi-Agent Systems
- AI and cyber security are underestimated: one in four companies does not provide training
Comments
Please log in to post a comment.