AI Helped Scammers Hit 12,000 Accounts

A criminal service called EvilTokens has been taken down after compromising 12,000 Microsoft accounts belonging to 10,000 organisations in a matter of months, according to Ars Technica.

EvilTokens launched in February via Telegram. For a $1,500 sign-up fee and $500 a month, subscribers got a complete fraud pipeline: an AI chatbot that read victims' inboxes, spotted trusted contacts and payment workflows, then drafted convincing impersonation emails designed to trick employees into transferring money. Selling the capability as a packaged monthly service is the same model as a legitimate SaaS product — price list, recurring billing, and all.

The platform exploited device code authentication, a legitimate sign-in method designed for TVs and devices without keyboards. The attacker sends the victim a code; the victim, thinking it is routine, enters it in their browser; the attacker's device gets authenticated. No stolen password required.

Microsoft, working with security firm SpyCloud and the UK's Metropolitan Police Service, seized 50 websites and 150 domains used to run the platform. Two men were arrested in the UK on suspicion of offences connected to EvilTokens. Victims spanned construction, healthcare, financial services, real estate, and higher education — mostly in the US, with significant numbers in Canada, the UK, Australia, India, and France.

What to do: Be sceptical of any authentication code request you did not personally trigger. If you receive a device code prompt out of nowhere, do not enter it — contact your IT support or the service directly to confirm it is legitimate.

Google has launched an early-preview feature for Pixel 11 that lets Gemini — Google's AI assistant — make phone calls to local businesses on your behalf, according to The Verge.

Tell it to make a reservation, check if a product is in stock, or reschedule an appointment. Gemini dials the number, navigates hold music and automated menus, and holds a conversation with whoever picks up. You watch a live transcript the whole time and can take the call over at any point.

Google trialled something similar last year, but that version did not let you jump in. This one does.

It is available now, but only for Gemini paid subscribers who have enrolled in the Phone by Google Public Beta on Pixel 11 in the US.

Honest read: Worth trying if you are on Pixel 11 and hate hold queues — but the narrow eligibility means most readers will be watching from the sidelines for now.

What it is: A new cloud memory architecture that lets an AI assistant remember things about you across devices, without giving Google — or anyone else — access to what it stores.

What's new: Until now, privacy-focused cloud AI was stateless — meaning it forgot everything the moment a session ended, like a conversation with someone who has no long-term memory. This adds persistent memory without moving the privacy line.

How it works: Your personal data is stored in an encrypted vault in Google's cloud. The cryptographic keys that decrypt it live only on your own devices. When an AI model needs your context, an end-to-end encrypted channel connects your device to a secure enclave (an isolated processing environment, sealed off from the rest of the cloud). The enclave decrypts just enough to handle the request, saves any new context, then re-encrypts immediately. According to Google, even its own engineers cannot access the contents during this process. The software running inside the enclave is published as a tamper-proof public record, so independent researchers can verify it has not been altered.

What it can't do: The independent audit covers the architecture, not every possible implementation. The privacy guarantee depends entirely on the cryptographic model working as described — and that claim has not yet been verified outside of Google's own published whitepaper and the single audit it has commissioned.

Who should care: Anyone using a Google AI assistant who has wondered where their personal context actually lives.

Try it / read it: Google DeepMind blog

If you build or run applications on OpenAI's API — the programming interface that lets software connect to OpenAI's models — GPT-6 now has an improved prompt caching system worth understanding.

Prompt caching works like this: when your app sends the same instructions or context repeatedly, the model stores a copy and reuses it instead of reprocessing from scratch. OpenAI is offering up to 90% off on those cached tokens (the chunks of text the model processes), with eligible shared content cached within a 30-minute window.

New this week: a Prompt Caching Dashboard to track hit rates over time, a diagnostics tool that tells you exactly why a cache miss occurred (for example, if your tool definitions changed), and explicit cache breakpoints so you choose what to preserve.

You can now also adjust reasoning effort between requests without breaking the cache — useful if some follow-up tasks need less processing power than the original.

Access the tools and guide at openai.com.

Worth knowing: This is an API-level feature — it affects developers building on GPT-6, not end users of ChatGPT directly.

Automated licence plate reader company Flock Safety faced a Senate Judiciary subcommittee hearing this week — without its CEO, who declined to testify, according to The Verge.

Flock's cameras scan plates, build location histories on vehicles, and share data across law enforcement agencies. Senators from both parties raised concerns: immigration enforcement access, tracking people seeking reproductive care across state lines, and a case where a Florida woman spent 13 days in jail after a false match.

No legislation was announced. If you live in the US, this is the policy debate to watch.