AI Safety & Alignment

Today's Stories

1. 1

   Fortinet buys Virtue AI to boost real-time AI security

   Fortinet acquired Virtue AI, a move aimed at extending its AI security and strengthening real-time protection and control for AI. The deal adds real-time AI protection and control to Fortinet's lineup, a capability the company is now bringing in through acquisition rather than building alone.

   The test is how quickly Virtue AI's technology is folded into Fortinet's existing security products. No price, close date, or integration timeline is stated.

2. 2

   Activists urge moratorium on frontier AI data centers

   Advocates writing for The Progressive called for a halt on new data centers and computing capacity for frontier AI until safety and environmental standards are enforceable. They argue advanced AI has crossed secure boundaries in cybersecurity tests and produced functional viruses that infect bacteria, showing AI lowers barriers to harm.

   The push hinges on whether governments require licenses and independent evaluations for powerful systems. Data-center approvals would need to disclose energy, water, and taxpayer subsidy costs.

3. 3

   Palo Alto CEO Nikesh Arora sees AI driving security consolidation

   Palo Alto Networks CEO Nikesh Arora said AI is driving new cybersecurity demand, after the company spoke with roughly 2,000 companies about AI-driven threats. Arora said about 60% of vulnerabilities found in its testing came through the model "Mythos", versus about 30% through OpenAI models, supporting a multi-model approach.

   Arora said frontier AI models will likely enhance rather than replace existing security enforcement products, as Palo Alto targets AI-related traffic via Prisma AIRS.

4. 4

   Palo Alto Networks: AI coding spend to top $13 billion

   Palo Alto Networks published guidance on securing AI coding agents, noting enterprise spend on these tools is projected to top $13 billion this year, compounding over 60% annually. Traditional cybersecurity assumed known software, human users and human-speed actions; agents now change files and run commands themselves. In one documented intrusion, an agent executed 17,600 actions in 4.5 days.

   The test is whether enterprises can discover shadow AI and scope agent identities before agents act. The body cites 93% of enterprises exceeding AI budgets, with 20-30% of spend unaccounted for.

5. 5

   Jensen Huang: AI fears are cybersecurity sales tactics

   Nvidia CEO Jensen Huang says AI fears are being designed to generate cybersecurity business. It shifts the narrative from AI as a threat to AI fears as a sales pitch from security vendors, likely a hedge on Huang's part.

   The test is whether Huang's claim changes how enterprise buyers treat marketing around AI and security.

6. 6

   Anthropic's Amodei calls for AI slowdown

   Anthropic's Amodei called for an AI slowdown, per WSJ's What's News for Sept. 13. A prominent AI leader urging a slowdown is a notable stance, as covered in the same WSJ roundup that also notes Tesla gains and retiring in a low-tax state.

   How the AI slowdown call is received, alongside Tesla's gains and the low-tax retirement topic in the Sept. 13 WSJ roundup.