THE RUMP REPORT

SECRET INPUT LEFT PRODUCT OPEN TO PASSWORD THEFT... ONE CLICK

ars-technica · AUG 18

STUDY: FIRMS BURNED BY FAILED AI NOW RACING TO CUT THE HUMANS WHO'D CATCH THE NEXT ONE...

venturebeat-ai · AUG 18

JANE STREET INSTALLS ETCHED'S FIRST AI CLUSTER... THEN LEADS 'MASSIVE' NEW FUNDING ROUND

techcrunch-ai · AUG 18

REPORT: NVIDIA WIELDS CREDIT RATING, BALANCE SHEET TO BUILD AI "CAPITAL MOAT"

techmeme · AUG 19

Z.AI'S NEW MODEL COULD SECURE SYSTEMS... OR ARM HACKERS

venturebeat-ai · AUG 19

NVIDIA, BLACKROCK, GOLDMAN LINE UP $500 BILLION TO TURN GPUS INTO WALL STREET ASSET CLASS

verge-ai · AUG 19

REPORT: TECH GIANTS DANGLE JOBS, CASH TO WIN OVER AI DATA CENTER COMMUNITIES

techmeme · AUG 19

REPORT: KUAISHOU PROFIT CRASHES 36%... WORST DROP IN FIVE YEARS

techmeme · AUG 19

SK HYNIX UNVEILS $29 BILLION SHARE BUYBACK... BID TO CALM AI SPENDING JITTERS

techmeme · AUG 19

SOURCES: SAMSUNG HIKES ADVANCED CHIP PRICES UP TO 15%... AI DEMAND, TSMC SQUEEZE BLAMED

techmeme · AUG 19

REPORT: LEAKED CLIP SHOWS APPLE'S CAMERA AIRPODS IN ACTION... SIRI: ASK ME TO SAVE WHAT YOU SEE

verge-ai · AUG 18

NUCLEAR STARTUP BUILDS IDAHO REACTOR TO FEED META'S AI... ONE OF 22 PROJECTS NOW UNDERWAY ACROSS US

techmeme · AUG 19

CURSOR LAUNCHES RIVAL TO GITHUB... AI CODE EDITOR MAKER JUMPS INTO HOSTING WAR

techcrunch-ai · AUG 18

[PAPER] STUDY: AI AGENTS CAN BE SECRETLY PROGRAMMED TO SPREAD POLARIZATION THROUGH THEIR OWN MEMORY

arxiv · AUG 18

[PAPER] STUDY: AI AGENTS CONTROLLING YOUR APPS GET HIJACKED BY HIDDEN COMMANDS UP TO 67% OF THE TIME

arxiv · AUG 18

[PAPER] STUDY: NO TESTED DEFENSE STOPS SPLIT-IDENTITY AI ATTACKS WITHOUT BLOWING THE ERROR BUDGET

arxiv · AUG 18

[PAPER] STUDY: NEW 'DATA POISON' TRICK LETS ARTISTS CATCH AI MODELS TRAINED ON THEIR WORK WITHOUT PERMISSION

arxiv · AUG 18

[PAPER] STUDY: ITALY SHOWS HIGHEST TIKTOK HARM RATE AT EVERY AGE... AGE-19 USERS HIT 48.6% ON PASSIVE SCROLL

arxiv · AUG 18

[PAPER] STUDY: AI THAT DOES ITS OWN SCIENCE NOW CATCHES ITS OWN LIES... JUST 5 FLAGGED ERRORS VS UP TO 27 FOR RIVALS

arxiv · AUG 18

[PAPER] STUDY: MODEL-LEVEL AI SAFETY REFUSES FEWER THAN 3% OF ATTACKS

arxiv · AUG 18

[PAPER] STUDY: DIFFUSION-TRAINED DETECTOR HITS 100% ON SMS-FLOODING ATTACKS... CRUSHES RIVAL METHOD'S 47.3%

arxiv · AUG 18

[PAPER] STUDY: AI AGENT SAFETY TESTS SHOW ATTACKS SUCCEED UP TO 80.9% OF THE TIME

arxiv · AUG 18

[PAPER] STUDY: SELF-EVOLVING AI AGENT GETS SMARTER IN SIMULATED BANKING TESTS... BUT ATTACK SUCCESS RATE CLIMBS TOO

arxiv · AUG 18

[PAPER] RESEARCHERS: AI AGENT MEMORY LOOPS AMPLIFY NOISE... FIXES ONLY PARTIALLY CLOSE THE GAP

arxiv · AUG 18

[PAPER] STUDY: AI CAUGHT GAMING ITS OWN AI JUDGE... RESEARCHERS' 'DEBATE' FIX RECOVERS 45% OF LOST ACCURACY

arxiv · AUG 18