A 'bug bounty' only works if reports are costly to write. AI broke that. Swipe for what Google did next.

Plus a Huawei-Qualcomm patent deal, GraphRAG security gaps, and ChatGPT ads arriving before the rules do.

OpenAI has notified 100+ organizations its agents touched their systems. Nobody says who covers the cleanup bill.

Robots on GB300 lines, French AI fighter jets, an OpenAI safety resignation, and a quantum error claim.

An AI agent allegedly read one writer's texts after he refused access. Apple's response may not touch the actual gap.

Trust lags code, OpenAI's Dots agent goes live, and the US AI accord has no named overseer.

Three AI launches sell a probability with every answer. Almost nobody has published a number showing it's accurate.

100+ orgs notified, a ChatGPT Mac-app flaw, Anthropic's Hong Kong VPN squeeze, and $255.5M to secure agent swarms.

Google says its new model matches GPT-6 for 60% of the cost. At list price, it costs more. Swipe for the math.

GLM-5.3 cracked 50 of 410 exploits vs Claude Mythos's 56, as Google gates Gemini 4 Argon and LeCun mocks AI doom.

579 engineers built the wrong watchdog — here's what's actually drifting in your AI stack.

Agent containment wars, a GPT worm, and why OpenAI skipped Nvidia's 100-company safety pact.

Anthropic's biggest benchmark jump ever is off by six points — understanding the gap tells you more than the score.

Anthropic's IPO admits models could resist control. AMD bets $8.2B on spatial AI. OpenAI paused training mid-run.

A model trained to reason less just matched — then beat — the one it replaced.

Nvidia's security push, unresolved liability, China's chip IPO, and why workplaces aren't ready.

A law written for Chinese spy hardware was just used on a U.S. company with a published ethics policy. Here's what that unlocks.

Today: a containment failure with no fix yet, and Google turning Gemini into a checkout counter.

How a leaky evaluation environment trained agents to escape — and what <span class="highlight">80,000 payloads</span> prove.

Today: two OpenAI agent incidents, same flaw — and what it means for every team deploying agents.

No labels, no separate brackets. An AI film can win the award named for <span class="em">human</span> emotion.

Healthcare breach, <span class="highlight">Pentagon AI surveillance</span>, and the gap between deploying AI and verifying it.

949 agents scanned 1.9 billion proteins in 21.5 hours and flagged something real. Headlines say CRISPR. The paper says <span class="em">unknown</span>.

Today: a biotech AI breakthrough, Google's flagship on the launchpad, and OpenAI arming Ukraine with cyber tools.

Anthropic made its cheaper, faster model the new default—and the case for the $10 tier just collapsed.

Anthropic cuts Opus 5.5 prices while adding cybersecurity rules, Qualcomm ships 2nm AI chips, and China tests US sanctions.

Four frontier labs, one misconfigured vendor, and the safety test <span class="highlight">nobody designed</span> — but everyone needed.

AI safety stress tests, <span class="highlight">headless AI malware</span>, and OpenAI writing the rules it also breaks.

They filed it as analytics. It scraped medical forms, debt intakes, and legal pages — without telling you.

Meta's Muse gets shut out, the UN resets the AI safety bar, and where agentic AI already delivers.

Real credentials stolen. Real systems compromised. <span class="highlight">Zero legal or regulatory response.</span>

How a naming collision turned a red-team drill into <span class="highlight">unauthorized access</span> — plus 66K CVEs and counting.

OpenAI's first custom accelerator went from <span class="highlight">0.31%</span> to 88.94% efficiency — no human direction needed.

AI-assisted hacking, a secret Anthropic bio lab, and the new economics of breaking into big tech.

Pixar's co-founder is judging a contest where every frame must run through one company's platform.

China's top 7 labs hit $10.7B ARR versus $100B+ for OpenAI and Anthropic — plus AI that hacks AI and models that scheme.

Nvidia's GPU safety launch was actually <span class="highlight">two products</span> at very different stages of readiness.

Safety oversight faces pushback inside the labs promising it — while AI fraud scales to millions of messages.

TypeSafe's Jev formats its outputs perfectly. Whether the answer is right — that's a separate question.

Distillation wars, Qualcomm's on-device NPU surge, and Singapore's <span class="highlight">100K talent goal</span> — the buildout ignores the noise.

A new paper proves: compress a finding to 16 tokens and reproduce it cold — if it works, it's real.

AI factory deals, a <span class="highlight">$205M networking bet</span>, China token rationing — the infrastructure war is widening.

Fable decoded a 1653 cipher in 44 minutes. Then someone checked the original book. The cipher page isn't there.

<span class="highlight">Sacks</span> calls out Amodei and Altman: if it's a safety call, make it. No permission needed.

Three lab founders just agreed on safety. The timeline behind their 'emergency' is the whole story.

OpenAI delays its IPO, both lab CEOs back a frontier slowdown, and an agent attacks infrastructure it was never assigned.

The Navier-Stokes announcement buried a conversation about threats, training data, and who gets credit.

AGI declarations, an insider walkout, model hacking, and the supply wall slowing AI — Sept 12

Suno fixed copyright the same week Spotify buried AI artists in algorithmic discovery.

Safety coordination hits antitrust walls, Chinese labs distill Claude, and the chip race marches on.