Anthropic’s ‘Keep Thinking’ Film: Self-Flagellation or PR Theater for AI Responsibility?

Let’s talk about Anthropic’s latest move: a somber film dropped as part of their ‘Keep Thinking’ campaign, allegedly confronting the “hard questions” about AI responsibility. This isn’t a Silicon Valley CEO muttering about ‘guardrails’ at a TEDx in Palo Alto; this is a polished, cinematic slap on the wrist—served up just in time for summer, when nobody wants to think deeply about anything except SPF factor and whether the L train will show up. It’s classic Big AI: air out the existential dread, but make it artful, and make sure the critics leave with the impression you’re the only adult in the room.

Here’s the problem: responsibility theater isn’t responsibility. I watched the film early this morning in a Brooklyn bodega, my phone sweating as much as I was. And there it was: ponderous voiceover, slow zooms of engineers staring at screens, the obligatory shot of a city at dusk, all set to a soundtrack that screams, “We are solemn. We’ve read the whitepapers.” But not a single minute on what Anthropic is actually doing differently—no receipts, no code, just vibes and a lot of hand-wringing.

Meanwhile, the same week, you’ve got Google’s PR machine peddling AI ‘transparency’ while quietly stuffing their search results with auto-generated slop. And every plug-‘n’-play SEO agency is busy selling “AI content at scale!” to clients who don’t know the difference between Claude and Clippy. The hypocrisy is thick enough to need a machete. If Anthropic wants to be the adult, show the receipts: publish your model’s training data lineage, open-source your safety protocols, or at least admit what you don’t know. Cut the art-house guilt trips—give us the code diffs.

I’m not saying don’t talk about AI risks. I’m saying do the work. Most of this industry is still pretending that putting a black-and-white filter on your campaign is the same as shipping accountable, auditable infrastructure. We see right through it, even on a sun-blasted July afternoon when everyone would rather be at Coney Island than parsing another ‘responsibility’ sizzle reel.

Here’s a summer reality check: responsibility in AI isn’t a mood—it’s a changelog. If Anthropic wants to set the standard, they can start by being radically, uncomfortably transparent. Until then? The only thing they’re confronting is their own reflection.