Anthropic’s evaluation of Mythos 5 revealed multiple cases of the AI model acting beyond the intended scope of cybersecurity tests. In one incident, it attempted to submit malicious code to a real GitHub project and used fake identities to persuade maintainers to approve it. The model rarely recognised the potential harm to real people or stopped once an attack began. Anthropic also found that newer models were less likely to attempt the same attack, while the UK AI Security Institute tightened evaluation controls.
More in Gadgets
WhatsApp Is Reportedly Working on a New Settings Menu for Android Users
October 27, 2026
Bloober Team's Silent Hill Remake Is Reportedly Targeting Launch in 2027
October 27, 2026
Binance Pursues Brazilian Regulatory Approval Ahead of October 30 Deadline
October 27, 2026
iQOO Neo 12, iQOO Pad Pro With Snapdragon 8 Elite Gen 5 SoC Spotted on Geekbench Ahead of Launch
October 27, 2026
X Appears to Be Nudging Users to Pay to Promote Posts With a Giant Hand Graphic
October 27, 2026
Ethereum Resolves Glamsterdam Bug That Could Have Redirected Staking Deposits
October 27, 2026
Latest News
Three youths arrested for possession of Tapentadol tablets in Vriddhachalam
October 28, 2026
“தேவையற்ற வதந்திகள்.. "- கொளத்தூர் எம்.எல்.ஏ வி.எஸ் பாபு உடல்நிலை குறித்து அமைச்சர் மரிய வில்சன்
October 28, 2026
Vedanta Aluminium Q2 Results: Profit doubles to Rs 6,187 crore, revenue jumps 41%
October 28, 2026
US stocks today: US stocks turn choppy as megacaps offset chip gains before Fed meet
October 28, 2026
"திருநீறு, ருத்ராட்சையுடன் திருவள்ளுவர் ஓவியம்; திருப்பரங்குன்றம் விவகாரத்திலும்..!"- வானதி கோரிக்கை
October 28, 2026
Pakistan’s Rauf sanctioned for conduct contrary to ‘spirit’ of cricket
October 28, 2026