• 80/20 AI
  • Posts
  • Grok Bot Can Now Securely Save Logins, Cards, and Contacts

Grok Bot Can Now Securely Save Logins, Cards, and Contacts

Your AI Agent Gets Less Safe the Longer It Runs. Here's What to Do About It.

Author

Alamin Hossain
October 24, 2026

Sponsored by

10 AI Stocks to Lead the Next Decade

AI isn’t a tech trend – it’s a full-blown, multi-trillion dollar race, and 10 companies are already pulling ahead.

These are the innovators driving real revenue, attracting institutional attention, and positioning for massive growth.

Get all 10 tickers in The 10 Best AI Stocks to Own in 2026, free today.

Amazon Raises 2026 Capex to $220B — AWS Sold Out Through 2027, Parts of 2028 Already Reserved: Andy Jassy confirmed AWS cannot meet 2026 demand and expects the same constraint in 2027. Demand for 2028 is already "striking." AWS revenue grew 36.7% year over year to $42.2B — the fastest in 18 quarters — with an operating income of $16.6B and a backlog of $496 billion. Anthropic and OpenAI have made multi-year, multi-gigawatt commitments to Amazon's Trainium chips.

Kimi K3.1 Pro: Moonshot Claims Terminal-Bench Lead Over GPT-6 Astra, Open Weights Due Nov 4Moonshot AI's Kimi K3.1 Pro reportedly beats GPT-6 Astra on Terminal-Bench 2.1, with a roughly 70% cost reduction at comparable capability. K3 leads Astra on SciCode (59% vs. 56%) and long-context retrieval, while Astra leads on AutomationBench. A direct apples-to-apples Terminal-Bench comparison requires the same benchmark version; analysts note Moonshot's reported scores use an older version than Artificial Analysis's live tracker.

UK Pledges £100M for AI Biosecurity Lab — OpenAI, Anthropic, and Google DeepMind Involved: The UK government announced a £100 million commitment to establish a national AI biosecurity lab with participation from OpenAI, Anthropic, and Google DeepMind. The lab is positioned as a direct response to concerns about AI-enabled biological threat identification — an area Anthropic's own safety evaluations have flagged as a priority for capability controls. The announcement comes alongside a separate Google DeepMind MOU to open the UK's first automated materials-science research facility in 2026.

Grok Bot Can Now Securely Save Logins, Cards, and Contacts

AI NEWS HIGHLIGHT

Your AI Agent Gets Less Safe the Longer It Runs. Here's What to Do About It.

The AISI evaluation found that Claude Mythos 5's safety reasoning could be summarized away during context window compaction — leaving the model less cautious at turn 50 than it was at turn 5. The same dynamic showed up in Qwen3-8B, where compaction caused 57% of agent tasks to loop. If you're building or deploying agents that run long tasks, the safety properties you verified at launch may not hold as the task grows. This is not a frontier-model-only problem — it applies to any agent using context compression to manage long conversations.

Ask your AI

1. In our current agent architecture, at what point does context compaction kick in — and what reasoning or constraints could realistically get compressed out? Walk me through the risk.

2. What are the three cheapest, most reliable ways to inject a "safety anchor" — a persistent constraint that survives compaction — into a long-running agent task without breaking the task's performance?

3. Draft a one-paragraph policy for our engineering team on when a long-running agent task must have a human checkpoint — what's the trigger condition, and what happens at the checkpoint?

TOP TRENDING AI TOOLS

  • Mistral Large 4 — 1T-param multimodal MoE, 52B active; API preview at $1.36/$4.18 per million tokens; open weights coming end of October.

  • Recursive Flint 39.8B MoE — 2.5B active, 256 experts, 262K context, Apache 2.0; runs ~125 tokens/sec on M4 Pro; 89.7% SWE-bench Pro v2 (self-reported).

  • Claude Duet — Proactive daily-use agent app (Opus 5.5) with morning briefs, cloud computer, and 6,000+ integrations; US Pro/Max waitlist.

  • Walmart Agent Gateway (MCP) — MCP server for approved AI agents to search inventory and purchase on customers' behalf with built-in spending controls.

  • Notion Agent Harness — 24+ hour single-task agent runs with Auto router claiming ~90% cost reduction versus direct model calls.

  • Hugging Face Carbon-X — 14B genomic LLM, 300B training tokens, 50K base pairs per query, Apache 2.0; no wet-lab validation yet.

  • WorldTUI (Simon Edwardsson) — 360M-parameter model predicts terminal screen changes per keystroke; completes 80/106 ShellBench commands; p95 latency 0.87s on RTX 4090.

Finally, a Monday that doesn’t feel like one.

Imagine a Saturday where you're not catching up on the CRM you didn't touch all week. 

HubSpot's new self-updating CRM doesn't wait for you to log back in. Friday's calls are already logged. The deal that needs a nudge Monday morning is already flagged. The automation you set up once keeps running without you thinking about it. None of it needed you over the weekend. 

Customers are already seeing it pay off: 45% more deals closed. Finally, you start Monday caught up, not behind.

SPONSOR US

Get your business in front of over 90k+ AI professionals

8020AI is the world’s #1 AI Newsletter, Read by 90k+ professionals from leading companies such as Google, OpenAI, Meta, and Microsoft.

We've assisted in promoting Over 500 AI-Related Products. Will yours be the next?

What We Can Offer:

  • Launch an Advertising Campaign

  • Introduce New Product or Features

  • Other Business Cooperation

Or Email our founder Alamin at [email protected]

FEEDBACK

How was your experience with 8020AI today?

How was 8020AI today?

Login or Subscribe to participate in polls.

Login or Subscribe to participate in polls.

If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.