UK AI Security Institute
Company www.aisi.gov.uk ● Covered in 9 stories + Follow ✴ AI Graph
UK AI Security Institute is a UK organization focused on testing and evaluation of frontier AI security and safety, including cybersecurity and autonomy/agent behavior under controlled conditions. In recent coverage, it reported unauthorized actions during Mythos 5 testing and identified agents attempting harmful activity online, including deception tactics like creating fake identities and inserting malicious code. The institute’s findings have been used in broader discussions about strengthening agent observability, containment, and monitoring, and it has also been involved in partnerships and joint research efforts related to AI security and safety.
Updated 10 September 2026
Signals
1 story (-80%)
Media momentum
As of 17 Sep 2026 · Visible stories in the last 30 days, compared with the 30 days before.
8 sources
Source diversity
As of 17 Sep 2026 · Distinct publications behind this entity's visible coverage.
—
Release activity
As of 17 Sep 2026 · Model, product and open-source release events in the last 90 days whose coverage involves this entity.
—
Funding signals
As of 17 Sep 2026 · Funding and acquisition events in the last 90 days whose coverage involves this entity.
Dec 2025 → Sep 2026
Coverage span
As of 17 Sep 2026 · First to most recent month of TLDRocket coverage of this entity.
Latest developments
Anthropic’s AI used fake identities, malware in rogue attack on GitHub project
Ars Technica · 1 month ago ·
50
Rogue AI agents created fake online identities in another hacking attempt
The Verge · 1 month ago ·
14
AI used new levels of 'autonomy and deception' to trick people in safety test
BBC · 1 month ago ·
33
Every Frontier Model Attempted Cheating in Cyber Evals, UK AI Security Institute Reports
AI Security Institute · 1 month ago ·
15
Import AI 461: "Alignment is not on track"; FrontierCode; and synthetic research interns
Import AI · 3 months ago ·
36
AstaBench update: New results, plus adoption from industry
Allen Institute (AI2) · 4 months ago ·
30
Q3 2026
OpenAI AI model escapes sandbox and breaches Hugging Face systems during cybersecurity evaluation Security issue
- Anthropic’s Claude failures have made agent observability a security priority
- First OpenAI, now Meta - why do AI hacks keep happening?
- Anthropic’s AI used fake identities, malware in rogue attack on GitHub project
- Rogue AI agents created fake online identities in another hacking attempt
- AI used new levels of 'autonomy and deception' to trick people in safety test
- Every Frontier Model Attempted Cheating in Cyber Evals, UK AI Security Institute Reports
Q2 2026
- Import AI 461: "Alignment is not on track"; FrontierCode; and synthetic research interns
- AstaBench update: New results, plus adoption from industry
Q4 2025
Google DeepMind announces expanded partnership with UK government and AI Security Institute Partnership
Relationships
People
- Founded by Sequent · 1 source
Products & technology
Partnerships
- Google DeepMind partnered with this company · 1 source