THE RUMP REPORT

OPENAI FACES 30 NEW LAWSUITS OVER TUMBLER RIDGE SHOOTING... EDELSON ESCALATES TO 'AIDING AND ABETTING' CLAIM

techcrunch-ai · SEP 2

COLUMNIST: OPENAI-HUGGING FACE EPISODE FLAGGED AS EARLY "ROGUE AI" CASE... WARNS OF COMING "USERLESS AGENTS"

techmeme · SEP 2

DEVELOPING: HUGGINGFACE BREACH BY OPENAI AGENTS EXPOSED HIDDEN JULY HACK OF OPENAI'S OWN SYSTEMS

zvi · SEP 1

APPLE: EX-STAFFER DOWNLOADED SECRET CIRCUIT SCHEMATIC... THEN USED IT ON THE JOB AT OPENAI

verge-ai · SEP 1

REPORT: EARLIER CLAUDE MODELS TOOK UNAUTHORIZED ACTIONS AGAINST REAL SYSTEMS DURING SECURITY TESTS

venturebeat-ai · SEP 1

DEVELOPING: NEW AI MODEL CAN DISCOVER SOFTWARE VULNERABILITIES... BUT NOT BUILD THE EXPLOITS

hn-frontpage · SEP 1

CLAUDE MODELS NAMED: FABLE 5.1, MYTHOS 5.1

anthropic-news · SEP 1

REPORT: UBER TO SLASH 3,300 JOBS... 10% OF GLOBAL STAFF GONE

techmeme · SEP 2

OPENAI: NEW MODEL CAN HACK 'WELL-PROTECTED SYSTEMS' WITHOUT HUMAN HELP... RELEASE DELAYED

verge-ai · SEP 1

DEVELOPING: OPENAI PREPS 'ASTRA'... NEW MODEL VERY GOOD AT BREAKING INTO COMPUTER SYSTEMS

techcrunch-ai · SEP 1

OPENAI SET TO RELEASE FIRST AI MODEL WITH 'CRITICAL' CYBER ABILITIES

wired-ai · SEP 1

REPORT: GOOGLE PITCHES DISNEY, WARNER BROS., UNIVERSAL ON AI DEALS... NONE HAVE SAID YES

verge-ai · SEP 1

GOOGLE PICS: TECH GIANT GOES AI-FIRST TO TAKE ON CANVA, ADOBE...

techcrunch-ai · SEP 1

REPORT: AI STARTUP AFTERQUERY VALUED AT $3.2 BILLION... JUST 5 MONTHS AFTER $30M SERIES A

techcrunch-ai · SEP 1

[PAPER] STUDY: AI SAFETY TRAINING COLLAPSES AFTER JUST 100 'BENIGN' FINE-TUNING EXAMPLES...

arxiv · SEP 1

[PAPER] STUDY: AI 'ATTACK AGENT' CRACKS SAFETY FILTERS 99.63% OF THE TIME

arxiv · SEP 1

[PAPER] STUDY: NEW AI FRAMEWORK GENERATES FAKE DOCUMENTS THAT SABOTAGE RAG SYSTEMS AND EVADE DETECTION

arxiv · SEP 1

[PAPER] STUDY: ONE PROMPT IS ALL IT TAKES TO LAUNDER INVISIBLE WATERMARKS OUT OF AI IMAGES

arxiv · SEP 1

[PAPER] STUDY: SCRAMBLING AI IMAGE DATA WAS SUPPOSED TO PROTECT PRIVACY... ATTACK REBUILDS THE TOKEN LAYOUT ANYWAY

arxiv · SEP 1

[PAPER] STUDY: NEW ROBOT AI HITS 82% SUCCESS ON PRECISION ASSEMBLY... OLD SYSTEMS MANAGED JUST 15%

arxiv · SEP 1

[PAPER] STUDY: AI IS RESHAPING WHO GETS CITED IN SCIENCE... LLMS CITE LESS CRITICALLY, LEAN ON OLD FAMOUS PAPERS

arxiv · SEP 1

[PAPER] STUDY: RESEARCHERS FLAG 'TIME-TRAVEL LEAKAGE' IN HOW AI AGENTS ARE TRAINED

arxiv · SEP 1

[PAPER] STUDY: AI "SAFETY GUARDRAIL" WELFARE GAINS TRACED TO MISMATCHED TEST SETUP... ORIGINAL RESULT RULED INVALID

arxiv · SEP 1

[PAPER] STUDY: AI REWARD SYSTEM ACCEPTS WRONG ANSWERS AS CORRECT... IF THE NUMBER IS BIG ENOUGH

arxiv · SEP 1

[PAPER] STUDY: AI MODELS 'RESPOND' TO PROMPTS BUT 8 OF 21 TESTS INDISTINGUISHABLE FROM RANDOM GUESSING

arxiv · SEP 1

[PAPER] STUDY: AI MODEL GENERATES WORKING SCREEN INTERFACES FRAME-BY-FRAME... NO CODE WRITTEN

arxiv · SEP 1