5 stories this week that change your decisions (Jun 1-7, 2026)

TL;DR Anthropic banned 832 accounts for AI-assisted attacks, with 84.4% using AI for defense evasion and 69% for capability development, and found agentic scaffolding, not the raw model, is what most uplifts attackers. Separately, researchers built a proof-of-concept worm driven by a local open-weight LLM that exploited 73.8% of a 33-host test network and replicated onto 61.8% with no frontier-model API calls. And Microsoft added seven agentic failure modes, the most common attack bypassing the human approval gate so the agent acts unchecked.

1. Anthropic scores how much AI uplifts real-world attackers

832 accounts banned. 84.4% used AI for defense evasion and 69% for capability development. Agentic scaffolding, not the raw model, is what most uplifts attackers, and MITRE ATT&CK has no IDs for autonomous execution.

2. LLM self-replicating worm

A local open-weight LLM powered a proof-of-concept worm that exploited 73.8% of a 33-host test network and replicated onto 61.8%, showing that adaptive AI-driven replication can work without a frontier model or API calls.

3. Microsoft adds seven failure modes for AI agents

After twelve months of red teaming, Microsoft updated its taxonomy with seven new agentic AI failure modes. The most common attack bypasses the human approval gate, so the agent acts unchecked. A single poisoned memory can survive into later sessions.

4. The AI innovation and security executive order decoded

Washington gets a free seat at vulnerability discovery and 30-day pre-release access to frontier models. Federal systems get patched first, and with NSA in the room, some flaws may be kept for offense rather than disclosed. Voluntary on paper, steered by federal spending.

5. MIT identified five AI risks with over a 10% chance of catastrophic outcomes

Even with pragmatic, cost-effective mitigations, five AI risks still carry over 10% odds of catastrophe, and all 24 stay above 5%. Those five are dangerous capabilities, weapons and cyberattacks, power centralization, inequality & unemployment, and environmental harm, with the first two highest at 21%.

Sources: