JadidJust-in AI Developments, Incidents & Deadlines
April 2026
SMTWTFS1234567891011121321415161718192021222324252627282930

AI news, Tuesday, April 28, 2026

6 stories

TLDR

  • OpenAI misses revenue and user targets (Apr 28) — WSJ bombshell triggers a sharp AI-sector selloff: Oracle -6%, Nvidia/AMD -3–5%, SoftBank -10%. CFO Sarah Friar reportedly at odds with Sam Altman over spending discipline and IPO timeline.
  • Microsoft Entra AI Agent ID role flaw disclosed (Apr 28) — Silverfort found that the "Agent ID Administrator" role—designed for AI agents—could be abused to take over arbitrary service principals. Patched Apr 9; public disclosure today.
  • Ineffable Intelligence raises record $1.1B seed (Apr 27) — DeepMind RLReinforcement learningTraining by trial and error, where the model takes actions and adjusts based on the rewards it gets.Kaelbling et al. 1996 ↗ pioneer David Silver launches a new lab aimed at RL-based superintelligence with zero human data. Backed by Sequoia, Lightspeed, Nvidia, and Google.
  • Snap cuts 1,000 jobs, cites AI (Apr 22) — Adds to the growing tally: ~48% of Q1 2026 tech layoffs are now explicitly AI-attributed.

Action Items

Things your engineering team should look at or do this week.

  • Audit Entra ID for Agent ID Administrator role scope — If your tenant has any service principals with the Agent ID Administrator role, verify the April 9 patch has been applied across all cloud environments. The pre-patch flaw allowed that role to take ownership of any service principal (not just agent identities), enabling tenant-wide takeover if the target SP held directory or Graph app permissions. Details at Silverfort's write-up.

  • Pressure-test OpenAI budget assumptions — Today's WSJ report (OpenAI missing monthly sales targets, CFO warning on data-center commitments) increases the risk of service-level changes, pricing adjustments, or API strategy shifts ahead of a volatile IPO process. If your team has production workloads or cost models built around OpenAI's current pricing/capacity, flag them for review. (CNBC | Fortune)

Security

  • Microsoft Entra ID — Agent ID Administrator scope overreach (Apr 28, patched Apr 9) Agents Infra — Silverfort disclosed that the built-in "Agent ID Administrator" role—introduced by Microsoft for managing AI agent identities—had a scope overreach: it could assign ownership over any service principal in the tenant, not just agent-related ones. An attacker with this role could take over high-privilege service principals (including those with directory roles or sensitive Graph permissions) and add credentials to authenticate as them. Microsoft patched silently on April 9 after responsible disclosure on March 1. Any attempt to assign non-agent SP ownership now returns a "Forbidden" error. Check that all tenants are on the patched version. (The Hacker News | Silverfort | SecurityWeek)

Industry & Business

  • OpenAI misses revenue and user targets — AI sector selloff (Apr 28) Trending — The Wall Street Journal reported that OpenAI fell short of multiple monthly sales and new-user targets in 2026, as Anthropic gained ground in coding/enterprise and Gemini expanded its share. Oracle (which holds a $300B, 5-year compute contract with OpenAI) shed 6%; Nvidia dropped ~5%, AMD ~3%, SoftBank (a major investor) ~10%. OpenAI CFO Sarah Friar is reportedly concerned the company can't sustain its data-center commitments on current revenue trajectories — a concern that pits her against Altman's bullish IPO push. Altman and Friar jointly called the report "ridiculous," but the market reaction was swift. OpenAI is projecting a $14B loss in 2026 with breakeven not until 2030. (CNBC | Bloomberg | Fortune)

  • Ineffable Intelligence launches with record $1.1B seed (Apr 27) Open Source — David Silver — UCL professor and former head of DeepMind's reinforcement learning team (AlphaGo, AlphaZero) — launched Ineffable Intelligence with a $1.1 billion seed round at a $5.1B valuation, the largest seed in European history. Co-led by Sequoia and Lightspeed with participation from Nvidia, Google, DST Global, and the UK's Sovereign AI Fund. The company's thesis: train a "superlearner" entirely through reinforcement learning from raw experience — no human-generated training data — to reach what Silver calls "first contact with superintelligence." A high-conviction RL bet from one of the field's most respected practitioners. (TechCrunch | CNBC | Bloomberg)

  • Snap cuts 1,000 jobs, explicitly cites AI (Apr 22) Hiring — Snap CEO Evan Spiegel cited AI reducing repetitive work when announcing 1,000 layoffs on April 22. The cuts follow a broader pattern: the tech industry laid off ~80,000 workers in Q1 2026, with approximately 48% of those cuts attributed by companies to AI-driven automation and efficiency gains. Combined with the Meta/Microsoft 20K cuts (Apr 24), the AI-labor displacement narrative is accelerating faster than most 2025 forecasts predicted. (CNBC | Tom's Hardware)

AI in Production

  • OpenAI IPO risk: CFO flags revenue-capex mismatch (Apr 28) Trending — See Industry & Business. The operational implication for engineering teams: if OpenAI's revenue trajectory forces cost-cutting or pricing recalibration ahead of its IPO, API users may face surprises. CFO Friar reportedly flagged that OpenAI has entered compute commitments (including the $300B Oracle deal) that its current revenue may not support. Altman wants to IPO in Q4 2026; Friar has privately questioned whether the company is ready. Both called today's WSJ story "ridiculous" — but neither disputed the underlying numbers.

Policy & Safety

  • White House AI Framework: federal preemption push continues (ongoing, Mar 2026) — The White House's March 20 National Policy Framework for AI recommends Congress preempt state AI laws that "impose undue burdens," aiming for a single national standard. Congress hasn't acted yet, but Colorado's AI Act (takes effect June 30, 2026) and California's CCPA automated decision-making rules (effective Jan 2027) are live clocks — especially for teams deploying AI in HR, lending, or hiring decisions. The GUARDRAILS Act (introduced Mar 20) seeks to block that federal preemption. Outcome unclear; compliance teams should plan for multi-jurisdiction rules through at least 2027. (Holland & Knight | Alston & Bird)

Type to search · ↑↓ to move · enter to open