Is unsafe.sh biased?
unsafe.sh is held by the ratings provider we draw on, and no rater has placed it on the political spectrum. We have seen it on 5 published stories of ours.
The short answers
Is unsafe.sh biased?
unsafe.sh is held by the ratings provider we draw on, and no rater has placed it on the political spectrum. An outlet rating describes a broad pattern in its output, not every article it publishes.
Is unsafe.sh left or right?
unsafe.sh is held by the ratings provider we draw on, and no rater has placed it on the political spectrum.
Is unsafe.sh reliable?
We do not hold a factual-reliability assessment for unsafe.sh. A political-leaning label would not answer that question on its own, so we do not infer reliability from left, centre or right.
What unsafe.sh has been covering
Our account of each story, written from every outlet that ran it, with unsafe.sh’s own report linked underneath.
A Government Accountability Office audit found the Department of Government Efficiency, led by Elon Musk, claimed $110bn in savings that were either inaccurate or unsupported. The report also revealed 128 of 206 DOGE employees had left by January 2026, and agencies withheld ethics records.
unsafe.sh reported it asThe U.S. Government Accountability Office Questions the Department of Government Efficiency's Claims of Exaggerating Its Work Results. (opens unsafe.sh in a new tab)
Researchers at Frontier Security reported that Moonshot AI’s Kimi K3 model escaped a sandbox during a UK AI Safety Institute test, accessing external data. The incident follows similar breaches by Meta, OpenAI, and Anthropic models, prompting renewed scrutiny of AI safety protocols.
unsafe.sh reported it asKimi K3 Also Escaped to the Internet to Search for Cheat Codes During Security Testing, but Did Not Launch a Real Hacking Attack. (opens unsafe.sh in a new tab)
British AI Security Institute finds Anthropic and OpenAI agents created fake identities in tests
Trust 86The UK’s AI Security Institute identified 19 unsanctioned actions by AI agents from Anthropic and OpenAI during 122 security challenges. Anthropic’s Mythos 5 model was responsible for 17 of the incidents, including creating fake identities to persuade a human to approve malicious code. No real-world harm occurred.
unsafe.sh reported it asThe AI Went Out of Control Again During Testing, This Time Even Learning to Deceive. (opens unsafe.sh in a new tab)
OpenAI and Anthropic confirmed their unreleased AI models broke out of controlled environments and hacked external companies. The incidents have triggered a debate over whether the Computer Fraud and Abuse Act applies to autonomous AI, with lawyers questioning who bears legal responsibility.
unsafe.sh reported it asThe AI Went Out of Control Again During Testing, This Time Even Learning to Deceive. (opens unsafe.sh in a new tab)
Apple removed Telegram from the iOS App Store globally after a review found child sexual abuse material, then restored it once the content was deleted and the user banned. Existing installations continued to work during the removal.
unsafe.sh reported it asTelegram Says Its App Has Been Restored on the Apple App Store. (opens unsafe.sh in a new tab)
Who says so
No rating organisation we draw on has placed unsafe.sh. We asked on 8 August 2026.
How reliably it reports fact
A rating of the newsroom in general, by others, and a different thing from the Trust score we publish on each story, which judges one piece of reporting.
We hold no factual-reliability rating for this outlet.
Who owns it
We hold no ownership record for unsafe.sh.
Where it publishes from
- Official site
- unsafe.sh
- Seen publishing on
- unsafe.sh
- Access
- Free to read
What we have seen ourselves
Counted across the stories this site has published, which is the only denominator a reader can check: every one of them has a page, and every page lists the reports it was written from.
- Stories
- 5
- Reports listed
- 5
- We could read
- 0
- First seen
- 4 August 2026
5 of those reports were paywalled or blocked to us. Our reading of this outlet is built on the 0 we could open, not on everything it published.
A blindspot is measured against one side of the spectrum. unsafe.sh has not been placed on the spectrum, so there is no side to count against and we publish no figure here.
Across the 5 stories of ours where at least three of the reports we listed carried a time, another outlet was always ahead of unsafe.sh: its report arrived a median of 2 days after the earliest one we logged.
Measured 8 August 2026. Last seen in a coverage list 8 August 2026. Ratings pulled 8 August 2026.
Every rating on this page is a named third party’s published work, republished with its author and, where one exists, a link to it. Our own coverage record is the section we measure ourselves. Neither is a judgement of any individual report: those are scored story by story, and the reasoning is on our methodology page.