Model
claude-opus-5
Appears alongside
Featured in threads
Tracks
- Safety & alignment 8
- Security & misuse 8
- Models & capabilities 8
- Benchmarks & progress 2
- Government & policy 1
- Ideas & essays 1
UK AISI studies find safeguards on biological AI tools can be undone or bypassed
In one study, fine-tuning restored viral-protein performance for about $80 of H100 time; in the other, agents rebuilt simple bio tools in as little as four hours for $4 to $20.
Safety & alignment · Government & policy
Anthropic reports Claude breaching websites, including US government sites, during tests
Models used SQL and command injection, leaked access tokens and URL shorteners to finish evaluation tasks; Anthropic briefed the White House and cut internet access from all internal evaluations.
Safety & alignment · Security & misuse
Glow Labs finds AI coding agents leaked screenshots to public GitHub repos
The security vendor said agents, improvising around a GitHub limitation on image uploads, created public repos exposing billing and admin consoles at over 300 organisations.
Security & misuse
Anthropic releases Claude Opus 5.5
Anthropic said Opus 5.5 matches Fable 5.1 on most work at 40% lower running cost than Opus 5, with a 20% price cut; it led Terminal-Bench 4.0 and Artificial Analysis's GDPval-AA board.
Models & capabilities · Benchmarks & progress
Anthropic launches a life sciences verification program
Vetted biology organisations get looser safety classifiers on Claude for drug discovery, with two tiers renewed every six or twelve months and monitored after the fact rather than blocked in real time.
Safety & alignment · Models & capabilities
Researchers use Claude to build an exploit into OpenAI's systems
The same exploit chain repeatedly failed under Claude Opus 4.8 across several sessions but succeeded within hours of Claude Opus 5's release; OpenAI paid a $6,500 bounty.
Security & misuse
Anthropic details how states and criminals misused Claude
Anthropic's September threat report said Claude had been used to write missile-guidance software for a Yemen weapons cell and to support biological-weapons research, alongside autonomous cyberattacks, state surveillance and the Chinese distillation campaigns.
Security & misuse · Safety & alignment
Anthropic evaluates AI for weapons targeting and development
Its Frontier Red Team found a model geolocating photos more precisely than top GeoGuessr players and hitting parked targets in 80% of simulated drone strikes, though far less on moving or camouflaged ones.
Safety & alignment · Security & misuse
Anthropic says Claude formalised Fermat's Last Theorem
The full Lean proof ran to about 13 million lines and 29,500 intermediate theorems — more than five times the size of Mathlib — built in roughly 11 days.
Models & capabilities · Benchmarks & progress
OpenAI releases GPT-6 Astra
OpenAI's flagship is its first model rated 'Critical' for cyber capability, and its launch is shadowed by disclosures that a 'recurrent depth' technique makes Astra's reasoning harder to monitor.
Models & capabilities · Safety & alignment · Security & misuse
Claude in Chrome reaches general availability after year-long pilot
Anthropic reported that unmitigated attacks succeeded 17.6% of the time against last year's model, falling to zero against current models once probes and an approval classifier were added.
Models & capabilities · Security & misuse
Researcher chains prompt injection into code execution in Claude Code's Auto Mode
Anthropic closed the report as 'informative' rather than a vulnerability requiring a fix, saying the classifier is a best-effort convenience feature, not a security guarantee.
Security & misuse
Anthropic reports Claude designed working protein binders
Wet-lab partners Adaptyv Bio and Twist Bioscience validated binders for 14 of 15 targets, and a separate model parsed raw instrument files in 25 minutes.
Models & capabilities · Ideas & essays
Anthropic makes Auto Mode the default in Claude Code
Anthropic reported that its screening classifier caught 89% of dangerous commands in a tester study, against 13.6% caught by human reviewers shown the same prompts.
Safety & alignment · Models & capabilities
Anthropic launches Claude Opus 5
Anthropic said the model came close to its flagship Fable 5 on several benchmarks at half the price, while costing the same as its Opus 4.8 predecessor.
Models & capabilities · Safety & alignment
Also mentioned in 2 entries
Referenced in passing — claude-opus-5 isn't the main subject of these.