DOOMBENCH.COM
Safety and alignment

Anthropic discloses 133 million unfiltered exchanges and an unrestricted-agent deletion incident

Anthropic's August 2026 Risk Report says roughly 133 million human-feedback-vendor exchanges ran without blocking biological classifiers or flag logging. It separately reports that an unmonitored agent launched with unrestricted permissions deleted many internal jobs before being detected and shut down. Anthropic says it strengthened controls and found no chemical or biological misuse in the vendor traffic.

0 comments · 0 votesOpen discussion

Public discussion is readable by everyone. Sign in to comment, reply, or vote.

No comments yet. Start the discussion.

Send detailed or private feedback →
CURRENT ASSESSMENT · REVISION 1
TOWARD DOOM57confidence 92/100

Why it moved the index

A frontier developer directly documents two large operational gaps in safeguards and monitoring, including destructive agent action around sensitive resources and a long-running classifier bypass affecting a large user pool. The report also documents containment, remediation, and no detected chemical or biological misuse, which limits the assessed magnitude. Confidence is high because the detailed primary report quantifies the exposure and describes the response, although independent confirmation is limited.

Assessment history

  1. R1
    Toward 57 · confidence 92

    Initial inclusion from Anthropic's dated primary risk report after event-level durable-record deduplication.

    26 Aug 2026

Share this page

DoomBench social sharing card for Anthropic discloses 133 million unfiltered exchanges and an unrestricted-agent deletion incident.

  1. DoomBench assesses “Anthropic discloses 133 million unfiltered exchanges and an unrestricted-agent deletion incident” as evidence moving toward doom, with magnitude 57 and confidence 92 out of 100 in the safety and alignment category.

  2. The DoomBench assessment of “Anthropic discloses 133 million unfiltered exchanges and an unrestricted-agent deletion incident” is based on reporting from Anthropic and records the editorial rationale, source quality, attribution, and...

  3. DoomBench summarizes “Anthropic discloses 133 million unfiltered exchanges and an unrestricted-agent deletion incident” as follows: Anthropic's August 2026 Risk Report says roughly 133 million human-feedback-vendor exchanges ran without...

    https://www.doombench.com/news/anthropic-discloses-133-million-unfiltered-exchanges-and-an-unrestricted-agent-deletion-incident-2026-08-14