• Practically AI
  • Posts
  • 🧠 AI to fight fraud, interview humans, and beat Gemini

🧠 AI to fight fraud, interview humans, and beat Gemini

Today in AI: OpenAI speeds up GPT‐5.2, imper.ai takes on deepfake fraud, and Anthropic trains Claude to interview humans about AI’s impact.

OpenAI is fast-tracking GPT‑5.2 to win back users lost to Gemini 3. But it’s not just a race for speed — it’s a reset on trust, reliability, and keeping people inside the ChatGPT ecosystem.

Meanwhile, deepfake fraud is getting harder to spot — so a stealth startup just raised $28M to fight it without looking at content at all. And Anthropic’s turning Claude into an interviewer to track how AI is really changing our jobs.

If you’re trying to make sense of where AI’s going (and what to actually use this week) — this is the one to read.

Three big updates. Two tools to try. One workflow you’ll actually use.

Let’s get into it.

3 Big updates

1) GPT-5.2 could drop as early as tomorrow

OpenAI is rushing to launch GPT-5.2 — possibly as soon as December 9. Internal reports suggest this release will focus on practical improvements: faster response times, better reliability, and smarter personalization.

Behind the scenes, OpenAI has paused major features (like its “Pulse” AI assistant and shopping agents) to double down on the core ChatGPT experience.

Why this matters:

✔ GPT-5.2 may outperform Gemini 3 on internal benchmarks
✔ The company has seen a 6% drop in users since Gemini 3’s launch
✔ Expect a clear focus on speed, tone, and better follow-through in chat

➡️ If you’re a ChatGPT Plus user, keep an eye out for a silent rollout.

2) imper.ai launches with $28M to fight deepfake fraud

Brooklyn-based startup Imper.ai just launched with $28M in funding from Redpoint Ventures and Battery Ventures — aiming to tackle deepfake impersonation attacks without analyzing content.

Instead, their AI detects “digital breadcrumbs” attackers can’t fake: device data, network signals, typing patterns — things that are hard to spoof.

It already works across Zoom, Teams, Slack, WhatsApp, and Google Workspace.

Why this matters:

✔ Deloitte estimates $40B in annual AI fraud losses by 2027
✔ Imper’s approach may sidestep deepfake content arms races
✔ Enterprises are already testing quiet deployment across internal tools

➡️ Think of it like a silent fraud detector for your communication stack.

3) Anthropic launches Claude-powered Interviewer

Anthropic released an AI Interviewer trained to speak with professionals and track how AI is changing their work. In a recent study, it analyzed 1,250+ interviews across industries and surfaced useful contrasts — like where people feel displaced vs. where they’re gaining speed and clarity.

It’s part of a broader research push alongside the launch of Clio, Anthropic’s public-facing research assistant.

Why this matters:

✔ Gives product teams real feedback on how AI is affecting users
✔ Helps companies detect friction points, fears, and unmet needs
✔ The model is being used internally to shape Anthropic’s roadmap

➡️ Want to run your own? Check out deepfeedback.ai — a public version is now available for builders.

2 Tools worth trying

SciSpace Biomedical
If you work with research or med-tech teams, SciSpace now offers an AI assistant trained to interpret biomedical literature, PubMed papers, and datasets — and explain them in plain English. Ideal for non-specialists who require clarity.

X Design
An AI-first interface tool that helps teams design faster with smarter suggestions, reusable components, and real-time feedback. Especially helpful for solo designers or small product teams that need to move quickly.

1 Workflow for the day

Use Stitch by Google to visualize AI workflows

Stitch is Google’s new AI tool that turns text prompts or rough sketches into working UI designs — complete with layout, styling, and even exportable code.

You can describe what you want (“a mobile screen with a task list and filter buttons”) and Stitch will generate a clean interface. You can then export it to Figma or copy the code directly into your dev environment.

What it’s great for:
✔ Rapid wireframes when you're short on time
✔ Frontend mockups before writing a single line of code
✔ Bridging the gap between product, design, and dev

Example: I asked Stitch to design a "File Upload" screen for a mobile app. Here’s what I got:

Check out the visual sample here → Stitch by Google

Calling the community

Got a workflow you use every week? A clever prompt that saves hours?
Reply with a quick description — we’ll feature the best ones next week.

Was this post useful?

Login or Subscribe to participate in polls.

Until next time,
Kushank @digitalSamaritan

Reply

or to participate.