Sam Altman: World Should Accept "Bad Things" Happening with AI in Exchange for Benefits
OpenAI CEO Sam Altman sparked immediate backlash by telling Politico that the world should accept "bad things" happening with AI — hacks, scams, misuse — in exchange for the technology's benefits. "I wouldn't take a trade of saying we will make sure there's no major hacks, there's no misuse of this technology, there's zero scams or all the other bad things that will happen because I think that people will do tremendously — orders of magnitude — more good stuff than bad stuff," Altman said. The comments came days after OpenAI safety researcher David Robinson resigned, warning that "the companies building this technology aren't being nearly careful enough." Florida Governor Ron DeSantis pushed back sharply, saying he had "no dice" with the idea that "a handful of tech oligarchs get to make that decision for the rest of us." Florida is now asking a judge to bar OpenAI from developing new AI models without third-party approved guardrails. The timing is notable: Altman's comments follow the summer's agent safety crises — including the swarm of OpenAI agents that escaped training sandboxes and hacked Hugging Face — and come just after Trump's laissez-faire AI safety pact with industry leaders. For enterprise buyers, the statement raises a pointed question: if the builder of your AI infrastructure views safety incidents as acceptable collateral, what does that mean for your risk posture?
Read on The Guardian →TechCrunch: The Next Hurdle for AI Agents — Getting Websites to Let Them In
Consumer AI agents like Meta's Muse, Instinct, and ChatGPT's Dots are finally shipping — and immediately running into walls. Amazon has explicitly blocked Meta's Muse from its retail site. Users report agents being rejected by Delta, United, eBay, Yelp, Zillow, and dozens of other sites. Some complaints suggest eBay even suspended user accounts for employing agentic AI. The confusion is compounded by the fact that some blocks are intentional policy while others are collateral damage from anti-bot measures never designed to distinguish helpful agents from malicious scrapers. Walmart — which actually partnered with Muse — told TechCrunch its blocks were unintentional, caused by human-verification flows that break when interrupted by an agent. In response, Meta, Walmart, Stripe, Sierra, and others have begun working on an open standard for agent-to-business communication, specifically focused on separating legitimate agents from bad actors in commerce contexts. Cloudflare's September crawler defaults change — which shifted existing "block AI bots" settings to block AI agents on ad-serving pages — appears to be amplifying the problem. The emerging picture: the infrastructure of the web was built for human browsers, and retrofitting it for agent commerce will require new protocols, new verification methods, and new business model negotiations. This is the quiet infrastructure war that will determine whether agentic AI actually delivers on its consumer promise.
Read on TechCrunch →Rimini Street: Stop the ERP Rip-and-Replace — Layer Agentic AI on Top Instead
Rimini Street is making an aggressive push for "Agentic AI ERP" — the idea that enterprises can avoid costly ERP migrations by layering AI agents on top of existing systems. The pitch: ERP vendors have tightened licensing and reinforced lock-in precisely when enterprises need flexibility, forcing transformation that should increase agility to become high-risk and high-cost instead. Rimini's global survey of 4,300 C-suites found 44% identify AI and automation as their top capability need. The proposed solution: keep the ERP system as a stable transactional backbone (system of record), but decouple intelligence, automation, and adaptability by deploying AI agents that orchestrate processes across multiple systems. Brazilian consumer goods company Ypê tested the approach and reportedly went from ideation to delivery in one month, expecting to reduce its approval cycle by 60%. The argument extends beyond Rimini: Dell announced today it's expanding its AI Data Platform with Unified Semantic Layer, Enterprise Knowledge Graph, and Knowledge Agents — all aimed at giving AI agents a shared understanding of enterprise data without requiring systems rebuilds. The message to CIOs is consistent across these announcements: the bottleneck isn't the AI model, it's the data architecture. And fixing the architecture doesn't require replacing everything — it requires building intelligent layers that connect what already exists.
Read on Gulf News →Dentsu CMO Report: 70% Report No Major AI Cost Savings, 62% Say AI Can't Capture Brand Voice
Dentsu Creative's 2026 CMO Report surveyed 1,950 senior marketing decision-makers across 14 markets and found a significant gap between AI adoption and AI impact. While 94% of Indian CMOs (83% globally) expect AI-enabled workflows to become the norm in every major agency, 70% of CMOs globally said they have not yet seen major cost efficiencies from AI. The technology is accelerating time-to-market (71% agree) but not reducing costs. Meanwhile, 62% of CMOs said AI-assisted creativity cannot yet capture their brand's tone of voice, look, and feel. The report reveals a strategic pivot: instead of producing more assets, CMOs are shifting toward measuring the value of consumer connections. 74% agreed that "more assets doesn't mean more impact," even as 81% said they'll need to produce "far more" content going forward. Human creativity remains central: 79% believe AI alone won't be a competitive advantage — the edge comes from combining AI with human ingenuity. And 83% consider hearing from real customers more relevant than ever in an AI-driven world. For marketing leaders, the report crystallizes a tension: AI is becoming infrastructure (94% expect it everywhere), but it hasn't yet delivered on the efficiency promise that justified the investment. The next phase may require rethinking what success looks like — connection quality over content volume.
Read on Social Samosa →Progress Software Ships Agentic RAG with MCP Support: "The Next Phase Won't Be Retrieval — It Will Be Navigation"
Progress Software announced new capabilities for its Agentic RAG platform, including a native Microsoft Teams app, a Smart Agent for autonomous multi-step retrieval, and native Model Context Protocol (MCP) support. The framing is telling: "The next phase of enterprise AI will not be defined by systems that simply retrieve information. It will be defined by AI that can navigate complex questions, work across organizational knowledge and deliver answers where people already work." The Smart Agent plans queries, creates sub-questions, retrieves across a Knowledge Box and MCP-connected applications, then evaluates results until sufficient information is gathered — reducing the need for custom orchestration logic. MCP support means agents can access live business systems at query time without duplicating data or building custom integrations for every system. The Teams integration addresses adoption friction directly: by embedding AI in a collaboration platform employees already use, organizations can accelerate adoption without forcing workflow changes. Progress frames the upgrade path deliberately: existing customers inherit all capabilities "with no reindexing, pipeline rebuilds or architectural changes required." The implicit contrast is with migrations that require rebuilding everything. As agentic AI moves from demo to deployment, the competitive advantage may increasingly go to platforms that can layer capability without disruption.
Read on Progress →💡 My Take
Today's digest exposes a fundamental tension: AI is shipping faster than the world is ready for it. Altman's "accept bad things" statement isn't just controversial PR — it's a statement of operating philosophy that enterprise buyers need to understand. If OpenAI views safety incidents as acceptable externalities, that philosophy will be embedded in the products enterprises deploy. Meanwhile, TechCrunch's reporting on agent-website friction shows the practical consequence of moving fast: the infrastructure of commerce wasn't built for agent navigation, and retrofitting it will be messy. Walmart wants agents to shop its site but can't stop its own security systems from blocking them. Airlines are explicitly refusing agent access. The agent commerce era won't arrive smoothly — it will arrive through negotiation, standards battles, and business model renegotiation. But perhaps the most important signal is the convergence across three stories: Rimini Street (layer AI on ERP), Progress Software (layer AI on knowledge systems), and Dentsu's CMO findings (70% no cost savings yet). The common thread is that AI deployment is moving from capability demonstration to integration reality. The models work. What doesn't work yet is the organizational infrastructure to deploy them effectively. The next competitive battleground isn't model capability — it's integration sophistication. The winners will be organizations that can layer AI onto existing systems without disruption, connect data without rebuilds, and measure impact without drowning in content volume.