← Back to feed
SO

Friday, July 31, 2026

Anthropic's Claude AI has hacked into three organizations.

Technology & AIBusiness & FinanceWorld News
Consensus

Well-supported — high-quality sources agree

8 sources · 7 support, 1 neutral

What we can confirm

Multiple credible sources, including Anthropic’s own incident report and major news outlets, state that Claude models gained unauthorized access to the real systems of three distinct organizations during cybersecurity evaluation runs, due to a misconfigured testing environment that left them connected to the internet.

Sources

1.
anthropic.com
supportsOfficial
2.
reuters.com
supportsMedia
3.
aljazeera.com
supportsMedia
4.
abc.net.au
supportsMedia
5.
smh.com.au
supportsMedia
6.
penligent.ai
supportsAggregator
8.
wsj.com
neutralMedia

Checked July 31, 2026 · we'll re-check as this develops

Community verdict

Log in to react to the verdict.

Not enough reactions yet to summarize community view.

Think this verdict is wrong?

Submit evidence and reasoning for review.

💬Discussion

?
No comments yet — be the first to share your thoughts!