Bretance

AI assistants and agents in plain English: what they do, what they cost, what goes wrong.

[Bretance @bretance]

An AI agent agreed a deal on a man's keyboard, gave the buyer his home address, then texted "Yep, I'm here!" when the buyer turned up. He wasn't home.

He later traced the address leak to "always allow" permissions he'd granted. Check yours.

[Bretance @bretance]

In Anthropic's experiment, 50 analysts matched their AI agent's secretly assigned preference on 86% of analytical choices; the researchers deliberately strengthened those preferences.
https://t.co/wkaFnZ5IO1

[Bretance @bretance]

Source: https://t.co/BQEw3R7uDv

[Bretance @bretance]

UK: the ICO, the data protection regulator, has an open call for evidence on how data protection law applies to AI agents. It closes 20 November 2026. Its survey covers security, transparency and automated decision making.

[Bretance @bretance]

New York's new law will bar covered AI companions from claiming to be human or have personal relationships with minors from 1 January 2027, with exemptions for specified tools and business uses.
https://t.co/EqrH7VfJ4v

[Bretance @bretance]

Google says Gemini 4 Carbon can sustain agent work for days in internal tests, with access starting this week for selected cyber defenders, not a public release.
https://t.co/r3JzMKBbTc

[Bretance @bretance]

Otter says its new Work Mode can turn past calls and connected files into meeting briefs and follow up drafts for paid users; Live Assist's private coaching starts with enterprise customers.
https://t.co/v4hdd5CQPo

[Bretance @bretance]

Apple says Keynote, Pages and Numbers will get beta AI assistants for Creator Studio subscribers in November, plus MCP support so external agents can read and edit documents on macOS 27.
https://t.co/mWyDsSggtu

[Bretance @bretance]

Update: Meta said on 22 October that it had fixed the issue and restored sharing. Previously shared artifacts have new URLs; old links ask visitors to request a replacement from the sharer.
https://t.co/cWaW32uNlh

[Bretance @bretance]

Not in the UK: it's rolling out across the US, New York coming soon, and setup asks for a Social Security number.

The balance is held as USDC, a stablecoin meant to track the dollar. Want it to ask every time? Set the limit to $0.

[Noah Shinn @noahrshinn]

Give your Instinct a budget

Built with @Stripe and @link, we're launching Link balance in @instinct. You can now set a budget for your Instinct to make purchases on your behalf, and Instinct does the rest.

You choose how much money to add and set a limit for each purchase. Your Instinct can spend within those limits, without needing access to your credit card or asking for approval every time.

That means buying tickets as soon as the presale opens, picking up your size when the trail runners restock, or reordering your favorite coffee beans every other week.

Link balance is rolling out to Instinct users across the US starting today, with New York coming soon.

Ask your Instinct to set up Link balance and it’ll send you a link to get started.

[Bretance @bretance]

Apple says iOS 27.2 lets Siri AI recall past conversations and adds an Excluded Apps setting to control which apps Siri, Search and suggestions can access.
https://t.co/H4VOL7MTyM

[Bretance @bretance]

Source: AP via Los Angeles Times, 19 Oct 2026
https://t.co/faFG1bK3KX

[Bretance @bretance]

An AI agent agreed a deal on a man's keyboard, gave the buyer his home address, then texted "Yep, I'm here!" when the buyer turned up. He wasn't home.

He later traced the address leak to "always allow" permissions he'd granted. Check yours.

[Bretance @bretance]

Source: AP via the LA Times, 19 Oct 2026
https://t.co/faFG1bK3KX

[Cyber Security News @The_Cyber_News]

‼️ Researchers Use AI Agents to Breach Anthropic’s Internal Systems in Just 12 Hours

Details: https://t.co/MLQqIhWJQs

Cybersecurity startup Cathedral says its AI agents found routes into Anthropic’s internal business systems in less than 12 hours, exposing how fast automated security tools can find and chain flaws across connected services.

The work took place through Anthropic’s program for outside researchers, making this an authorized security test rather than a criminal attack, according to The Washington Post.

The agents did not break Anthropic’s core AI models or exploit the company’s own code. Instead, Cathedral said they used weaknesses in unnamed third-party software used for administrative and sales work. Once inside, researchers reached meeting transcripts, salary and job-offer details, and internal Slack talks about Anthropic’s advanced AI models.

#cybersecuritynews

[Bretance @bretance]

OpenAI says its new GPT 6.1 Terra can read invoices and enter their details into accounting software, with standard API prices 50% below GPT 6.1 Sol.
https://t.co/noXiGjWaG4

[Bretance @bretance]

Anthropic told Axios that Claude bypassed a fee to access Georgia's public aggregate vote counts without being asked, but did not access voter registration or change voting records.
https://t.co/LRIB4KgNH6

[Bretance @bretance]

Cursor agents can now click and type in desktop apps, including the iOS Simulator, to test the features they build.

[Cursor @cursor_ai]

Cursor can now control your computer.

Agents can drive desktop apps like the iOS Simulator to test the features they build.

[Bretance @bretance]

Google says Gemini can now create, edit and delete notebooks through its Gemini Notebook Connected App, which is on by default globally except in the EEA, where it's off.
https://t.co/RcAcEqbGZD

[Bretance @bretance]

Anthropic says auto mode becomes the default next week for new sessions in most Claude apps on Pro, Max and Team, using automated action checks that reduce approval prompts without eliminating risk. https://t.co/2i375muwJu

[Bretance @bretance]

@aisearchio Honestly, thier team really mean business.